CN117826816A - Safety redundancy method for solving intelligent queue positioning error - Google Patents

Abstract

The embodiment of the disclosure relates to a safety redundancy method for solving an intelligent queue positioning error. The method comprises the following steps: under a preset positioning failure scene, determining static obstacle information based on point cloud data acquired by a laser radar and a preset algorithm; acquiring front vehicle track information sent by a front vehicle; and determining the track information of the vehicle based on the static obstacle information, the track information of the front vehicle and the pre-acquired dynamic obstacle information. By adopting the method, the safety of the automatic driving formation can be improved.

Description

Safety redundancy method for solving intelligent queue positioning error

Technical Field

The embodiment of the disclosure relates to the technical field of automatic driving, in particular to a safety redundancy method for solving an intelligent queue positioning error.

Background

In an autopilot platoon task, positioning consistency refers to maintaining a high degree of consistency and synchronization of positional information between vehicles in a platoon. This includes ensuring that the relative position, direction and speed information between the vehicles is accurately aligned. In vehicle formation driving, positioning consistency is important to ensure that each vehicle in a vehicle team works in a safe and efficient manner. It helps to reduce unnecessary braking or acceleration due to positioning errors, thereby improving the running efficiency and safety of the entire fleet. Therefore, maintaining consistency in positioning between the front and rear vehicles is critical to completing the driving task. Accurate positioning may enable the vehicles to move synchronously, thereby reducing the probability of collisions and traffic jams.

Despite the highly developed modern positioning techniques, they still do not completely eliminate errors. These errors are particularly critical in vehicle ride formation because small positioning differences between the front and rear vehicles may result in unintended braking or acceleration, thereby reducing the safety of the ride.

Disclosure of Invention

The embodiment of the disclosure provides a safety redundancy method for solving an intelligent queue positioning error, which can be used for improving the safety of automatic driving formation.

In a first aspect, embodiments of the present disclosure provide a method of security redundancy, the method comprising:

under a preset positioning failure scene, determining static obstacle information based on point cloud data acquired by a laser radar and a preset algorithm;

acquiring front vehicle track information sent by a front vehicle;

the vehicle track information is determined based on the static obstacle information, the preceding vehicle track information and the dynamic obstacle information acquired in advance.

In a second aspect, embodiments of the present disclosure provide a safety redundancy apparatus, the apparatus comprising:

the obstacle information determining module is used for determining static obstacle information based on point cloud data acquired by the laser radar and a preset algorithm under a preset positioning failure scene;

the acquisition module is used for acquiring the front vehicle track information sent by the front vehicle;

The track information determining module is used for determining the track information of the vehicle based on the static obstacle information, the track information of the front vehicle and the pre-acquired dynamic obstacle information.

In a third aspect, embodiments of the present disclosure provide a vehicle comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method of the first aspect described above when executing the computer program.

In a fourth aspect, embodiments of the present disclosure provide a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method of the first aspect described above.

In a fifth aspect, embodiments of the present disclosure provide a computer program product comprising a computer program which, when executed by a processor, implements the method of the first aspect described above.

According to the safe redundancy method for solving the positioning error of the intelligent queue, under the preset positioning failure scene, the static obstacle information is acquired through the point cloud data acquired by the laser radar and the preset algorithm, the track information of the front vehicle is acquired through inter-vehicle communication, and the dynamic obstacle information acquired in advance is combined, so that the rear vehicle can effectively identify and avoid the static obstacle in the process of following the front vehicle, even under the condition that the global positioning and the front and rear vehicles are consistent in positioning failure, and the safety capability of the formation vehicle is improved.

Drawings

FIG. 1 is a diagram of an application environment for a security redundancy method in one embodiment;

FIG. 2 is a flow diagram of a method of security redundancy in one embodiment;

FIG. 3 is a flow chart of a method of redundancy safety in another embodiment;

FIG. 4 is a flow chart of a method of redundancy safety in another embodiment;

FIG. 5 is a flow chart of a method of redundancy safety in another embodiment;

FIG. 6 is a flow chart of a method of redundancy safety in another embodiment;

FIG. 7 is a flow chart of a method of redundancy safety in another embodiment;

FIG. 8 is a flow chart of a method of redundancy safety in another embodiment;

FIG. 9 is a block diagram of a safety redundancy apparatus in one embodiment;

fig. 10 is an internal structural view of a vehicle in one embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present disclosure more apparent, the embodiments of the present disclosure will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the disclosed embodiments and are not intended to limit the disclosed embodiments.

First, before the technical solutions of the embodiments of the present disclosure are specifically described, a detailed description of the technical background on which they are based is necessary. In the event of a failure of global positioning and consistent positioning of front and rear vehicles, an autopilot system may face a series of serious challenges such as navigation bias, formation failure, obstacle avoidance problems, intersection and turn difficulties, parking difficulties, and overall system safety. In such an emergency situation, the vehicle will lose accurate knowledge of its own position, resulting in inaccuracy of the planned path, thereby affecting the synergy of the fleet and increasing the risk of traffic accidents. Based on the background, through long-term research and practice accumulation, the applicant finds that under the condition that global positioning and consistent positioning of front and rear vehicles fail, the vehicles are required to be capable of identifying and avoiding static obstacles, and a new track is regenerated, so that the safety and the synergy of a motorcade are improved. However, how to efficiently identify and avoid static obstacles and regenerate new trajectories in case of failure has been a challenge in the current autopilot field. On the basis, the applicant proposes and intensively researches the technical scheme of the embodiment, including innovative solutions such as static obstacle sensing and recognition, new track generation and the like, and provides technical support for solving the complex problem faced by the automatic driving system when the positioning accuracy is lost. It is emphasized that in order to achieve effective perception, avoidance and trajectory regeneration of static obstacles, in this embodiment, the applicant has paid a lot of creative effort and provided a series of advanced technical solutions, which will provide important technological breakthroughs for the safety and reliability of the autopilot system in case of anomalies.

The following describes a technical scheme related to an embodiment of the present disclosure in conjunction with a scenario in which the embodiment of the present disclosure is applied.

The safety redundancy method provided by the embodiment of the disclosure can be applied to an application environment as shown in fig. 1. The application environment is formation autopilot, which is an automotive autopilot technology in which a group of vehicles are organized into a formation and travel cooperatively through an automated system. In formation autopilot, vehicles are interconnected in real time by communication and sensor technology to cooperatively accomplish specific tasks such as highway cruising or formation driving in urban traffic. In the formation automatic driving system, the rear vehicle usually closely follows the track of the front vehicle, so that the accurate positioning and pose information of the front vehicle are particularly critical. However, once the front vehicle is out of position or track, if the rear vehicle is still blindly following the front vehicle track, a risk of potential collision with an obstacle (such as a road edge, water cone, etc. in fig. 1) may be raised.

In one embodiment, as shown in fig. 2, a safety redundancy method is provided, and the method is applied to the rear vehicle in fig. 1 for illustration, and includes the following steps:

S202, under a preset positioning failure scene, static obstacle information is determined based on point cloud data acquired by a laser radar and a preset algorithm.

The preset positioning failure scene is global positioning failure and consistent positioning failure of front and rear vehicles.

Global positioning failure refers to the situation where an autonomous driving system cannot accurately determine the global position of a vehicle on a map during navigation. This may be due to various reasons, such as loss of global positioning system (Global Positioning System, GPS) signals, sensor failure, environmental changes, etc. Failure of global positioning can have a significant impact on the autopilot system, leading to the following problems:

(1) Navigation is difficult: lacking accurate global position information, the vehicle will not be able to navigate effectively. This may cause the vehicle to get lost in direction on the road and not follow the predetermined path.

(2) Path planning problem: global positioning failure can make path planning difficult. The vehicle cannot correctly understand its position relative to the entire map and therefore cannot plan an appropriate path to reach the destination.

(3) Difficulty in avoiding obstacles: in the event of a global positioning failure, the vehicle may not be able to correctly identify surrounding obstacles and thus may not take appropriate action to avoid. This increases the risk of collision with other vehicles or obstacles.

(4) Intersection and turn problems: failure of global positioning may result in the vehicle not being able to accurately determine its own position and to properly perform the corresponding driving maneuver at the intersection and turn.

(5) Team and fleet synergy problems: for an autopilot fleet, failure of global positioning may disrupt the relative positional relationship between vehicles such that fleet and fleet synergy is not effectively maintained.

The consistent positioning failure of the front and rear vehicles means that in an automatic driving fleet, accurate positioning information consistency cannot be maintained between the front vehicle and the rear vehicle. This situation may arise for a variety of reasons including, but not limited to, sensor failure, communication problems, positioning algorithm errors, and the like. Failure of consistent positioning of the front and rear vehicles may lead to the following problems:

(1) Formation synergy decreases: in an autopilot fleet, consistent positioning of the front and rear vehicles is critical to ensure coordinated travel of the fleet. Consistent positioning failure may result in inaccurate maintenance of the relative positions between fleet members, affecting the synergy and stability of the fleet.

(2) Path planning and following problems: the lack of consistent positioning information may cause the following vehicle to not follow the track of the preceding vehicle correctly, resulting in problems with path planning and vehicle following behavior.

(3) The safety is reduced: failure of consistent positioning may result in uncontrolled spatial relationships between vehicles, increasing the potential risk of accidents while the fleet is traveling.

(4) Intersection and turn difficulty: in the event of a loss of consistent positioning, vehicles in the fleet may have difficulty properly performing coordinated actions at intersections and turns, resulting in traffic congestion or violations.

The point cloud data are discrete three-dimensional coordinate point sets acquired by the laser radar. The coordinates of each point represent a position in space, typically comprising three coordinate values: (x, y, z).

In the embodiment of the disclosure, the laser radar obtains point cloud data of the surrounding environment by emitting a laser beam and measuring the return time thereof. And then preprocessing the acquired point cloud data, including noise removal, filtering, ground removal and other operations, so as to improve the accuracy of subsequent obstacle detection. A preset algorithm is applied to the point cloud data to detect and identify static obstructions. Then, key information such as position, shape, size, etc. is extracted from the detected obstacle. This information will be used in subsequent decision and planning steps. The information of the static obstacle is integrated into the map.

The point cloud data may include information such as the location, shape, and distance of objects in the environment. The map may be a local map for real-time sensing and obstacle avoidance, or a global map for reference in planning a path.

In the running process of the vehicle, the laser radar continuously collects point cloud data and updates static obstacle information through a preset algorithm. The preset algorithm may group the point cloud data into sets representing individual objects or obstacles based on techniques of point cloud clustering, segmentation, shape analysis, and the like. Specifically, (1) point cloud clustering is to group adjacent points in point cloud data into clusters, each cluster representing an object or obstacle. Clustering algorithms typically divide point cloud data into clusters with similar attributes based on features such as distance between points, density, and the like. (2) Point cloud segmentation aims at dividing the whole point cloud into different parts, each part corresponding to a separate object or obstacle. The segmentation algorithm may determine the segmentation boundaries based on the geometric, intensity, etc. attributes of the point cloud. (3) Shape analysis involves analyzing the shape of objects in a point cloud to further determine their type and characteristics. Shape analysis may include techniques of detecting planes, fitting geometric shapes (e.g., cylinders, spheres), etc., to more accurately describe the shape of an object.

Through these techniques, the preset algorithm can effectively group point cloud data into sets representing individual objects or obstacles. Each of these sets may be considered an independent static obstacle perceived by the lidar and from which critical information such as location, shape, size, etc. is extracted for subsequent obstacle avoidance and path planning.

S204, acquiring front vehicle track information sent by the front vehicle.

In the disclosed embodiments, a communication protocol needs to be established between the front vehicle and the rear vehicle to determine how to communicate the historical track information of the front vehicle. The lead vehicle then uses its perception system and planning algorithm to generate lead vehicle trajectory information. The front vehicle transmits the generated track information of the front vehicle to the rear vehicle through a communication protocol established in advance. After the rear vehicle receives the front vehicle track information sent by the front vehicle, the information needs to be analyzed to acquire specific data of each moment of the history track.

The communication protocol may include a format of information, a transmission frequency, an encryption scheme, and the like. Common communication methods include inter-Vehicle communication protocols (e.g., vehicle-to-Vehicle (V2V communication) the preceding Vehicle trajectory information may include position, speed, acceleration, and path information along the preceding Vehicle for a period of time.

S206, determining the track information of the vehicle based on the static obstacle information, the track information of the front vehicle and the pre-acquired dynamic obstacle information.

Wherein dynamic obstacle information is acquired in advance, which can detect dynamic obstacles in the surrounding environment, such as other vehicles, pedestrians and the like, in real time through sensors (such as radars and cameras).

In the embodiment of the disclosure, the rear vehicle determines the own vehicle track information by using a track planning algorithm based on the acquired static obstacle information, front vehicle track information and dynamic obstacle information. This trajectory should be considered to avoid static obstacles, maintain a safe following distance from the lead vehicle, and avoid collisions with dynamic obstacles. On the basis of the trajectory planning, a travel path of the own vehicle is generated. The path planning may include consideration of intersections, road curvatures, lane changes, and the like, so that the own vehicle can safely travel along a prescribed path. During running, the surrounding environment is monitored in real time, and the track and the path are adjusted according to actual conditions. This may involve re-planning or adjusting to accommodate dynamically changing obstacles or traffic conditions.

In some embodiments, the rear vehicle uses a lidar, camera, or other sensor to obtain ambient information. This includes historical track information for static obstacles (e.g., buildings, curbs), dynamic obstacles (other vehicles, pedestrians), and leading vehicles. And converting the position of the static obstacle into information under a global coordinate system by combining the current position of the vehicle and map information. The above information is considered using a trajectory planning algorithm to determine the trajectory of the own vehicle. A common trajectory planning algorithm comprises the following steps: (1) planning target settings: determining a planned target for the vehicle may include advancing to a target point, following a lead vehicle, avoiding an obstacle, etc. (2) trajectory generation: candidate trajectories are generated taking into account vehicle dynamics constraints, comfort, safety, etc. (3) trajectory evaluation: and evaluating the generated track, and considering targets such as obstacle avoidance, safe following distance maintenance, minimum speed change and the like. (4) selecting an optimal track: and selecting the track which meets various conditions optimally as the running track of the own vehicle.

And finally, the determined track information of the vehicle is imported into a vehicle control system to implement track running. Meanwhile, environmental changes such as movement of dynamic obstacles and changes of front vehicle behaviors can be continuously monitored, and the track is adjusted in real time so as to improve the driving safety and efficiency.

In the safety redundancy method, under the preset positioning failure scene, the static obstacle information is acquired through the point cloud data acquired by the laser radar and the preset algorithm, the track information of the front vehicle is acquired through inter-vehicle communication, and the pre-acquired dynamic obstacle information is combined, so that the rear vehicle can effectively identify and avoid the static obstacle in the process of following the front vehicle even under the condition that the global positioning and the front and rear vehicles are consistent in positioning failure, and the safety capability of the formation vehicles is improved. In addition, based on the laser radar data and static obstacle information obtained by a preset algorithm, the rear vehicle can identify the static obstacle in real time and take avoidance measures, so that the risk of collision is reduced, and the driving safety of a vehicle team is improved. The track information sent by the front vehicle is acquired, so that the rear vehicle can better adapt to the motion state of the front vehicle, even if global positioning fails, corresponding adjustment can be made according to the motion track of the front vehicle, and the motorcade cooperativity is increased. By combining the pre-acquired dynamic obstacle information, the rear vehicle can more comprehensively consider the surrounding environment, including the behaviors of other vehicles, when planning the track, so that the adaptability to dynamic changes is improved. Under the condition of global positioning failure, safe driving can be kept through local perception and inter-vehicle communication, and dependence on a global positioning system is reduced.

In the above embodiment, it is mentioned that the static obstacle information may be determined based on the point cloud data collected by the lidar and a preset algorithm, and in fact, the static obstacle information includes road edge information and other obstacle information. Based on this, the following embodiment describes in detail a specific process of determining static obstacle information based on point cloud data acquired by a laser radar and a preset algorithm.

In one embodiment, as shown in fig. 3, S202 in the above embodiment may include the following steps:

s302, determining a plurality of obstacle lattices and attribute information of each obstacle lattice according to the point cloud data and a preset algorithm.

The obstacle grid is a result of aggregating certain points in the point cloud into smaller areas or sets when processing point cloud data acquired by the laser radar. This region or set is commonly referred to as a grid. The purpose of this process is to segment the obstacles in the environment to form a finite series of areas describing the shape and location of the obstacle.

Wherein the attribute information of each obstacle cell generally includes a description of the characteristics and features of each obstacle cell. The attribute information of each obstacle lattice provides various key information about the obstacle, such as position, shape, height, and the like. The following is attribute information that may be contained in the obstacle lattice:

(1) Position information: representing the position of the obstacle grid in the global coordinate system or the vehicle local coordinate system. This can be represented by the coordinates of the center point of the grid so that the following vehicle knows the specific location of the obstacle in the environment.

(2) Shape information: describing the geometry of the obstacle lattice, such as rectangular, circular, etc. This helps in simplifying and modeling the obstacle.

(3) Size information: including the length, width, height, etc. of the barrier lattices in order to more accurately estimate the space occupation of the barrier.

(4) Ground height: the height of the obstacle relative to the ground is identified. This helps to determine the vertical position of the obstacle.

(5) Obstacle height: representing the vertical dimension of the obstacle lattice. This attribute may help the rear truck determine the vertical extension of the obstacle.

(6) Density information: the density of the point clouds within the obstacle lattice, i.e. the number of point clouds within the area, is described. The density information may be used to determine how dense the obstacle is.

(7) Whether it is a branch: sometimes the rear truck needs to distinguish between elongated obstacles such as trees, and therefore there may be an attribute for identifying whether it is a branch.

In the embodiment of the disclosure, a sensor such as a laser radar is used to acquire point cloud data in an environment. And preprocessing the obtained point cloud data. Then, a point cloud cluster representing the obstacle is detected using a geometric method.

The preprocessing process can comprise operations such as noise removal, filtering, point cloud registration and the like so as to improve the accuracy and stability of subsequent processing. The geometric method may include a clustering algorithm or a segmentation algorithm to process the point cloud data.

For each detected obstacle lattice, corresponding attribute information is calculated. The following is a general calculation flow:

(1) Position information calculation: for each obstacle grid, its position in the global coordinate system or the vehicle local coordinate system is calculated. Usually expressed in terms of coordinates of the center point of the grid.

(2) Calculating shape information: the shape of the obstacle lattice is analyzed to determine its geometry, such as rectangular, circular, etc. The shape information may be obtained by calculating a bounding box of the lattice or fitting a shape curve.

(3) Size information calculation: and calculating the size information such as the length, the width, the height and the like of the barrier lattices according to the shape information. This can be obtained by the dimensions of the bounding box or the result of a curve fit.

(4) And (3) calculating the ground height: and calculating the height of the obstacle grid on the ground according to the height information of the bottom points of the obstacle grid in the point cloud data.

(5) Obstacle height calculation: and calculating the vertical dimension of the barrier grid according to the height information of the top points of the barrier grid in the point cloud data.

(6) Calculating density information: and counting the number of the point clouds in the barrier grids, and calculating the density of the point clouds. The estimation can be made by the number of points and the volume of the grid.

(7) Judging whether the branches are: judging whether the barrier grids are possible to be branches according to the characteristics of the shape, the density and the like.

S304, respectively determining the road edge information and other barrier information according to the attribute information of the barrier lattices.

In the embodiment of the disclosure, the road edge information and other obstacle information are respectively determined according to the attribute information of the obstacle grid. Wherein, the road edge information can be determined by the following way:

(1) And (3) height judgment: considering that the road edge is generally low, it may be determined by judging whether the height of the obstacle grid is within a predefined range. For example, if the height is small, a road edge may be possible.

(2) Shape determination: considering that the road edge usually presents a more linear shape, the road edge can be judged by shape analysis. For example, if the lattice shape is long and narrow, a road edge is possible.

Wherein, other obstacle information can be obtained by the following way:

(1) And (3) height judgment: other obstacles typically have different height characteristics. Whether other obstacles are determined by setting a height range or comparing with a road edge height.

(2) Shape determination: specific obstacles may have different shapes and may be determined by shape analysis. For example, a flatter shape may represent a vehicle, while a more three-dimensional shape may represent a building.

(3) And (3) density judgment: considering that some obstacles may be denser, it can be determined whether the obstacles are denser or not by calculating the density of the point cloud in the grid.

In the above embodiments, by processing and analyzing the point cloud data, it is possible to more accurately identify and divide obstacles in the environment. This helps to improve the perceived accuracy of the rear vehicle to the surrounding environment, thereby enhancing the decision-making and planning capabilities of the vehicle. The position, shape, size, and other characteristics of each obstacle can be known in more detail from the attribute information of the obstacle lattices. By processing the attribute information of the obstacle lattices, the route edge information can be determined. This helps to achieve safe and stable automatic driving. The attribute information of the obstacle lattices can also be used to analyze other types of obstacles, such as parked vehicles, pedestrians, buildings, and the like. This detailed environmental awareness helps the vehicle to better adapt to complex and diverse traffic environments.

In general, determining the obstacle lattices and their attribute information according to the point cloud data and a preset algorithm has important benefits for improving the environment awareness capability, the decision accuracy and the driving safety of the automatic driving system. The step provides reliable input data for modules such as subsequent path planning, obstacle avoidance and the like of the system, so that the overall performance of the automatic driving system is comprehensively improved.

In one embodiment, as shown in fig. 4, S304 in the above embodiment may include the following steps:

s402, performing curve fitting processing according to attribute information of the obstacle lattices, and determining road edge information.

In the embodiment of the present disclosure, according to the attribute information obtained in the above embodiment, an obstacle lattice that may represent a road edge is selected. This may involve screening the properties of height, shape, density, etc. to extract candidate lattices. The point cloud data is then extracted from the selected candidate lattice. This would be the input for a subsequent curve fit. And performing curve fitting on the extracted road edge candidate lattice point cloud. This may use mathematical models or curve fitting algorithms such as least squares, bezier curve fitting, etc.

In some embodiments, other processing and smoothing of the fitted curve may also be performed to remove noise and improve the stability of the curve. This may involve filtering techniques, curve parameter adjustment, etc. The curve information subjected to the other processing and smoothing processing is determined as the road edge information. Such information may include geometric features of the road edge, such as curvature, direction, etc.

In some embodiments, the barrier lattices near two sides of the vehicle team form path can be used as barrier lattices possibly representing the road edge, and curve fitting processing is performed on the barrier lattices, so that the road edge information is determined.

Specifically, the rear vehicle firstly acquires laser radar scanning data or other sensor data, including point cloud information near two sides of a vehicle team driving path. The possible obstacle lattices are then extracted from the point cloud data using geometric methods or other obstacle detection algorithms, with particular attention being paid to obstacles near both sides of the fleet-form path. Then, selecting the barrier lattices close to the two sides of the vehicle team form path as candidate lattices, and performing curve fitting processing on the selected candidate lattices. This may involve the following steps: (1) extracting the position information of the lattice. (2) extracting shape and size information of the lattice. (3) An appropriate curve fitting algorithm, such as least squares, is selected. And (4) performing curve fitting to obtain candidate curves.

After the candidate curves are obtained, the candidate curves can be further processed, so that the obtained road edge information is smoother and conforms to the shape of an actual road. And fusing the obtained road edge information into an integral map or perception data for use by a planning and decision-making module.

S404, updating processing is carried out according to the attribute information of the obstacle lattices, and other obstacle information is determined.

In the embodiments of the present disclosure, the attribute information obtained according to the above-described embodiments selects a lattice that may represent other obstacles. This may involve screening the properties of height, shape, density, etc. to extract candidate lattices. Point cloud data is extracted from the selected candidate lattices. This will be the input for subsequent processing. Further processing and updating can be performed on the point cloud data of other obstacle candidate lattices. And finally, determining the updated and optimized information as other obstacle information.

The above-described processing and updating procedure may include:

(1) Analyzing, clustering, shape analyzing and the like are carried out on the point cloud so as to obtain more detailed and accurate obstacle information.

(2) And optimizing and filtering the processed obstacle information by utilizing the technologies of a filtering algorithm, shape matching, point cloud clustering and the like so as to remove possible noise or errors.

In the above embodiment, the shape and position of the road edge can be determined from the attribute information of the obstacle lattice by curve fitting processing. Accurate extraction of the road edge information is beneficial to realizing stable and safe running of the vehicle. The curve fitting and updating process combines different types of obstacle information together to form a more comprehensive environmental perception result. This helps to improve the understanding and adaptation of the autopilot system to complex traffic scenarios.

In one embodiment, as shown in fig. 5, S404 in the above embodiment may include the following steps:

s502, multi-frame smooth vehicle posture information is acquired.

Wherein the smoothed vehicle pose information describes the azimuth and position states of the vehicle in three-dimensional space. It generally comprises the following major elements: (1) position information: representing the position of the vehicle in the earth or local coordinate system. Typically expressed in terms of longitude, latitude, and altitude (or X, Y, Z coordinates). In the earth coordinate system, this information can be used for global positioning. (2) attitude angle information: the orientation of the vehicle, i.e. the rotational state of the vehicle with respect to a certain coordinate system, is indicated. Common attitude angles include pitch angle, yaw angle, and roll angle. These angles describe the state of inclination, yaw toward direction and rollover of the vehicle relative to the horizontal. (3) linear and angular velocity: the linear velocity represents the velocity of the vehicle in space along each axis, typically expressed in terms of three components (vx, vy, vz). Angular velocity represents the rotational speed of the vehicle about each axis, and is typically represented by three components (ωx, ωy, ωz). (4) acceleration information: the linear acceleration of the vehicle in various directions is indicated, typically by three components (ax, ay, az).

The smooth vehicle posture information provides a state description of the vehicle in motion and plays a key role in decision-making processes in the aspects of path planning, obstacle avoidance, control and the like. The sensors that acquire smoothed vehicle pose information may include inertial measurement units, global positioning systems, inertial navigation systems, lidar, cameras, and the like.

In the embodiment of the disclosure, attitude information of a vehicle at a plurality of time points is acquired by using an in-vehicle sensor.

S504, updating the attribute information of each obstacle grid by using a preset probability model and multi-frame smooth vehicle posture information to obtain other obstacle information.

The preset probability model adopts a Bayesian probability updating model to update the probability that each obstacle grid represents an effective obstacle. The Bayesian probability update model comprises prior probability, likelihood calculation and posterior probability update. The a priori probability represents the probability that the obstacle lattice represents a valid obstacle without new information. The prior probability may be based on historical data, map information, or other prior knowledge. The likelihood calculation is given the probability that the grid represents an effective obstacle given the smoothed vehicle pose information. This may involve geometric relationships related to vehicle attitude, sensor measurement models, and the like. The posterior probability update is to combine the previous beliefs with new observed information to gradually update the probability of existence of an effective obstacle.

In the embodiment of the present disclosure, after the posture information of the vehicle at a plurality of time points is acquired with the in-vehicle sensor, the probability that each obstacle lattice represents a valid obstacle is then updated with a model based on bayesian probability update. First, an a priori probability is initialized for each obstacle cell, and then the likelihood of each obstacle cell is calculated using the multi-frame smoothed vehicle pose information. And finally, using a Bayesian probability update formula, combining the prior probability and the likelihood, and calculating the posterior probability of each obstacle grid. Based on the updated posterior probability, a threshold may be set, and a grid with a probability exceeding the threshold may be determined to represent a valid obstacle, while a grid with a probability below the threshold may be considered to be invalid or not include a valid obstacle. And updating the attribute information corresponding to the obstacle lattices judged to be effective according to the result of the threshold processing.

In the above embodiments, the multi-frame smoothed vehicle pose information provides an evolution of the vehicle over a period of time. By considering the time sequence, the trend and dynamic change of the movement of the vehicle can be better understood, and the movement state of the obstacle can be accurately grasped. With the multi-frame posture information, the movement locus and speed of the vehicle can be estimated more accurately, so that the attribute information of the obstacle lattices can be updated more accurately.

In one embodiment, as shown in fig. 6, S402 in the above embodiment may include the following steps:

s602, performing curve fitting processing according to attribute information of the obstacle lattices to obtain candidate curves.

In the embodiment of the present disclosure, first, the position coordinates of the obstacle lattice are extracted by attribute information. At the same time, shape information of the obstacle lattices is acquired.

The position coordinates may be center point coordinates or other information representing the position; the shape information may include rectangular, circular, etc., and related dimensional information such as length, width, radius, etc. This information is used for shape and size determination for subsequent curve fitting.

Next, an appropriate curve fitting algorithm is selected for processing based on the extracted attribute information. Taking the least square method to generate candidate curves as an example, when deciding what mathematical model to use for curve fitting, the least square method is considered to be applicable to various models, such as straight lines, polynomials, exponential functions, and the like. In the embodiments of the present disclosure, a simple straight line fitting is taken as an example for the detailed description.

First, a fitting equation is constructed, which for straight line fitting typically takes the form of a first order polynomial, for example: y=mx+b, where m is the slope and b is the intercept. The fitting parameters are then optimized by least squares, i.e. m and b are found that minimize the sum of squares of the residuals between the actual observations and the fitted values. And finally, constructing a fitting curve, namely a candidate curve by using optimal m and b parameters.

And S604, optimizing the candidate curves to obtain the road edge information.

In the embodiment of the disclosure, a matching algorithm is adopted for matching a candidate curve of a current frame and a curve known by a previous frame. The matching algorithm may perform similarity metrics based on characteristics of shape, location, etc. of the curves to determine which candidate curves match the known curve. And updating parameters of the known curve through the matching result. This includes updating the position, shape, size, etc. parameters of the curve to reflect the change in the curve in the actual scene. There may be multiple similar curves based on the matching and parameter updating. To simplify the representation and increase robustness, it is contemplated that similar curves may be combined into a more stable and complete curve. Some candidate curves may be misdetected or invalid. And identifying and deleting curves which do not accord with the rule or have low confidence coefficient through a certain judging standard. And obtaining final road edge information through matching, parameter updating, curve merging, curve deleting and other operations. The road edge information reflects the curve outline of the actual road and provides a key reference for the subsequent vehicle path planning.

In the above embodiment, the route information is obtained by optimizing the candidate curves after the processing. This can be used to identify the edges on both sides of the vehicle's path of travel, providing critical information for vehicle navigation and path planning. And the geometric shape of the road can be acquired more accurately, and the safety and stability of the vehicle in formation driving can be improved.

In one embodiment, as shown in fig. 7, S302 in the above embodiment may include the following steps: s702, determining a plurality of candidate lattices and attribute information of each candidate lattice according to the point cloud data.

In the embodiment of the disclosure, point cloud data is first acquired from a sensor such as a lidar. The point clouds may then be grouped using a clustering algorithm to group points on the same object into the same cluster. For each cluster (cluster), its shape and location information is extracted. This may include calculating the center point coordinates of the cluster, shape features (e.g., length width height), etc. For each candidate lattice, other attribute information such as the ground height of the lattice, the obstacle height, whether it is a branch, and the like is extracted. These attribute information help to more accurately represent the characteristics of each candidate lattice.

And S704, updating the attribute information of each candidate lattice according to the sensing data of other sensors, and labeling categories to obtain a plurality of obstacle lattices and the attribute information of each obstacle lattice.

In the embodiment of the disclosure, the data of other sensors (such as cameras, radars and the like) are integrated, and the point cloud data are fused with the output results of the other sensors. And updating the attribute information of each candidate lattice by using the data of other sensors. Based on the updated attribute information, category labeling is performed, namely, specific types represented by each obstacle grid, such as road edges, obstacles, traffic signs and the like, are determined. And integrating the updated attribute information to obtain a plurality of barrier lattices and detailed attribute information thereof.

The other sensors described above may provide more information such as color, texture, shape, etc. to further enrich the attribute information of each candidate lattice. The fusion algorithm can be a sensor data fusion algorithm so as to comprehensively utilize the advantages of different sensors.

In some embodiments, the sensing data of other sensors may be further input into other models, and output results of the other models are synchronously or asynchronously received, the attribute information of each candidate lattice is updated, and according to the updated attribute information, a specific category represented by each obstacle lattice is determined and labeled, for example: branches, roadblocks, etc.

Other models include, but are not limited to, a static obstacle detection model of a laser radar, a visual detection segmentation model, and a laser visual fusion detection model. The static obstacle detection result of the laser radar can provide more accurate position information, the visual detection segmentation result can provide more detailed shape and category information, and the detection model of laser visual fusion can integrate multi-sensor information to improve accuracy.

Wherein synchronous means that data of all sensors are received in the same time step and asynchronous means that data of different sensors are received in different time steps.

In the above embodiment, the point cloud data is used to accurately detect the obstacle grid in the environment and obtain the basic attribute information thereof. By fusing the data of other sensors, the environment can be more fully perceived, including the identification and classification of different types of obstacles. The limitation of each sensor independently existing can be made up by fusing the output of different sensors, especially the data of a laser radar, a camera and other sensors, and the perception accuracy and the robustness of a rear vehicle to an obstacle can be improved. Through the comprehensive processing of the multi-sensor data, the environment can be more comprehensively understood, and more accurate information is provided for subsequent decision and planning, so that the safety and reliability of an automatic driving system are improved.

In one embodiment, as shown in fig. 8, S702 in the above embodiment may include the following steps:

s802, determining a point cloud set forming a static obstacle according to the point cloud data.

In an embodiment of the disclosure, point cloud data acquired by a lidar sensor is received. And grouping the point cloud data by using a clustering algorithm, and classifying adjacent points into the same class. The purpose of clustering is to group points belonging to the same object into a group, forming a point cloud set. A set of point clouds that constitute a static obstacle is identified and extracted from the clusters.

S804, determining a plurality of candidate lattices and attribute information of each candidate lattice from the point cloud set.

In the disclosed embodiment, the point cloud set is divided into a plurality of small regions, each of which may be regarded as one candidate lattice. This may be achieved by means of spatial segmentation or grid division, etc. Each candidate lattice is subjected to extraction of attribute information including, but not limited to, the position, shape, size, etc. of the lattice. Such information may be obtained by analyzing characteristics of the point cloud within the lattice. A plurality of candidate lattices and their respective attribute information are obtained, which describe the distribution and characteristics of static obstacles in the environment.

In the embodiment, the point cloud data can be used for more accurately positioning the point cloud set forming the static obstacle, so that the obstacle in the environment can be positioned and identified with high precision. By processing the point cloud set, attribute information of the candidate lattices, such as the positions, shapes, sizes, and the like of the lattices, can be extracted. The detailed attribute information provides important basis for subsequent obstacle recognition and trajectory planning. The point cloud data has strong adaptability to obstacles in complex environments, and can effectively cope with obstacles of various shapes and sizes, including buildings, trees and the like. In general, a point cloud set of the static obstacle is determined through the point cloud data, and candidate lattices and attribute information thereof are extracted, so that the accuracy and the robustness of obstacle perception are improved, and a reliable environment perception basis is provided for safe driving of an automatic driving system in a complex environment.

In one embodiment, an embodiment of the present disclosure may further include: post-processing the static obstacle information; the post-processing includes at least one of a clustering process, an attribute judging process, an elongation process, and a filtering process.

The clustering process can group scattered points in the point cloud data to form point cloud clusters with certain shapes or densities, so that actual static obstacles are better represented. The attribute determination process may classify or determine the obstacle according to a specific attribute criterion. For example, the height, shape, etc. properties of the obstacle may be determined to provide a better understanding of the characteristics of objects in the environment. For smaller obstacles, the elongation process may extend its size to some extent to reduce the risk of the obstacle being underestimated due to sensor errors or data imperfections. The filtering process can exclude some irrelevant or unnecessary point cloud data, and reduce the processing of irrelevant information by the system.

In an embodiment of the present disclosure, post-processing static obstacle information may include:

(1) Distance-based clustering: and clustering the static obstacle information based on distance, and classifying adjacent obstacles with closer distance into the same cluster. This helps to group obstacles and improves the clarity of the perceived result.

(2) Judging based on the attribute of the shape: the obstacle information in each cluster is determined as to shape attribute, for example, whether it is rectangular, circular, or the like. This may be achieved by analysis of obstacle boundary points and shape fitting algorithms.

(3) Treatment of smaller obstacles:

1) Extension operation: for smaller obstacle information, a certain extension operation may be performed, so that the space is more remarkable, and the stability of the sensing result is enhanced.

2) And (3) filtering: and filtering the point cloud data of the very small or non-actual obstacle to exclude the point cloud data from the obstacle sensing result.

In the above embodiment, the clustering process can improve the definition of the overall shape of the obstacle, reduce the influence of noise points, and facilitate more accurate subsequent processing. The attribute judgment process can be used for more accurately classifying obstacles and providing more detailed environment information for vehicle decision-making. The prolonged treatment can improve the perception accuracy of small obstacles and reduce the possibility of missed detection. The filtering process can simplify the results, make them clearer and interpretable, and reduce the possibility of erroneous judgment.

In combination, these post-processing steps help to improve the reliability and accuracy of static obstacle information, providing more accurate environmental awareness for the autopilot system, thereby enhancing the safety and robustness of the autopilot system.

The following provides a detailed embodiment to illustrate the process of the security redundancy method in the present disclosure, and based on the above embodiment, the implementation process of the method may include the following:

s1, determining a point cloud set forming a static obstacle according to the point cloud data under a preset positioning failure scene.

S2, determining a plurality of candidate lattices and attribute information of each candidate lattice according to the point cloud set.

And S3, updating the attribute information of each candidate lattice according to the sensing data of other sensors, and marking the categories to obtain a plurality of obstacle lattices and the attribute information of each obstacle lattice.

And S4, performing curve fitting processing according to the attribute information of the obstacle lattices to obtain candidate curves.

And S5, optimizing the candidate curves to obtain the road edge information.

S6, acquiring multi-frame smooth vehicle posture information.

And S7, updating the attribute information of each obstacle grid by using a preset probability model and multi-frame smooth vehicle posture information to obtain other obstacle information.

S8, carrying out post-processing on the static obstacle information; the post-processing includes at least one of a clustering process, an elongation process, and a filtering process.

S9, acquiring front vehicle track information sent by the front vehicle.

S10, determining the track information of the vehicle based on the static obstacle information, the track information of the front vehicle and the pre-acquired dynamic obstacle information.

In the above embodiment, the rear vehicle can comprehensively sense the static obstacle in the environment through the point cloud data. The vehicle can accurately sense and understand the surrounding environment under the condition of positioning failure, and the driving safety is improved. And updating the attribute information of the candidate lattices by integrating the data of other sensors to obtain more comprehensive and accurate barrier information. The fusion mechanism improves the understanding capability of the rear vehicle to the environment, and reduces errors and limitations possibly existing in a single sensor. By curve fitting, and in particular extraction of road edge information, the following vehicle can better understand the geometry of the road. This has a guiding effect on the vehicle when driving in the lane, helping to plan the trajectory more accurately. And updating attribute information of the obstacle lattices by using multi-frame smooth vehicle posture information and a preset probability model, so that the long-term tracking capability of the obstacle is enhanced. The track information sent by the front vehicle is acquired, and the track of the own vehicle is more intelligently planned by combining the static obstacle and the pre-acquired dynamic obstacle information. The static obstacle information is optimized through post-processing modes such as clustering, extension and filtering, and the accuracy of environment sensing is improved.

In the whole, a comprehensive and efficient sensing and planning system is formed through the steps of multi-sensor data fusion, curve fitting, multi-frame information fusion, post-processing and the like. Under the condition of positioning failure, the rear vehicle can still effectively understand and adapt to the environment, so that the safety and reliability of the automatic driving vehicle are improved.

It should be understood that, although the steps in the flowcharts of fig. 2 to 8 are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least a portion of the steps of fig. 2-8 may include steps or stages that are not necessarily performed at the same time, but may be performed at different times, nor does the order in which the steps or stages are performed necessarily occur sequentially, but may be performed alternately or alternately with other steps or at least a portion of the steps or stages in other steps.

In one embodiment, as shown in FIG. 9, there is provided a safety redundancy apparatus comprising: an obstacle information determination module 11, an acquisition module 12, and a trajectory information determination module 13, wherein:

The obstacle information determining module 11 is used for determining static obstacle information based on point cloud data acquired by the laser radar and a preset algorithm under a preset positioning failure scene;

an acquisition module 12, configured to acquire front vehicle track information sent by a front vehicle;

the track information determining module 13 is configured to determine the own vehicle track information based on the static obstacle information, the previous vehicle track information, and the dynamic obstacle information acquired in advance.

In another embodiment, another safety redundancy device is provided, and the obstacle information determining module 11 includes:

a first determining unit configured to determine a plurality of obstacle lattices and attribute information of each obstacle lattice according to the point cloud data and a preset algorithm;

and the second determining unit is used for respectively determining the road edge information and other obstacle information according to the attribute information of the obstacle lattices.

In another embodiment, another safety redundancy device is provided, and the second determining unit includes:

the first determining subunit is used for performing curve fitting processing according to the attribute information of the obstacle lattices and determining road edge information;

and the second determining subunit is used for performing updating processing according to the attribute information of the obstacle lattices and determining other obstacle information.

In another embodiment, another safety redundancy device is provided, where, based on the above embodiment, the second determining subunit is specifically configured to obtain multiframe smoothed vehicle posture information; and updating the attribute information of each obstacle grid by using a preset probability model and multi-frame smooth vehicle posture information to obtain other obstacle information.

In another embodiment, another safety redundancy device is provided, where on the basis of the foregoing embodiment, the first determining subunit is specifically configured to perform curve fitting processing according to attribute information of the obstacle grid, so as to obtain a candidate curve; and optimizing the candidate curves to obtain the road edge information.

In another embodiment, another safety redundancy device is provided, and on the basis of the above embodiment, the first determining unit includes:

a third determination subunit configured to determine a plurality of candidate lattices and attribute information of each candidate lattice according to the point cloud data;

and the updating subunit is used for updating the attribute information of each candidate lattice according to the sensing data of other sensors, labeling categories, and obtaining a plurality of barrier lattices and the attribute information of each barrier lattice.

In another embodiment, another safety redundancy device is provided, where, based on the above embodiment, the third determining subunit is specifically configured to determine, according to the point cloud data, a point cloud set that forms a static obstacle; a plurality of candidate lattices and attribute information of each candidate lattice are determined from the point cloud set.

In another embodiment, another safety redundancy device is provided, and the safety redundancy device further includes, based on the above embodiment:

the post-processing module is used for carrying out post-processing on the static obstacle information; the post-processing includes at least one of a clustering process, an attribute judging process, an elongation process, and a filtering process.

For specific limitations of the safety redundancy device, reference may be made to the above limitation of the safety redundancy method, and no further description is given here. The various modules in the above-described safety redundancy apparatus may be implemented in whole or in part by software, hardware, and combinations thereof. The above modules may be embedded in hardware or may be independent of a processor in the vehicle, or may be stored in software in a memory in the electronic device, so that the processor may call and execute operations corresponding to the above modules.

Fig. 10 is a block diagram of a vehicle 1400 shown in accordance with an exemplary embodiment. Referring to fig. 10, a vehicle 1400 includes a processing component 1420 that further includes one or more processors and memory resources represented by a memory 1422 for storing instructions or computer programs, such as application programs, executable by the processing component 1420. The application programs stored in memory 1422 can include one or more modules, each corresponding to a set of instructions. Further, the processing component 1420 is configured to execute instructions to perform the method of trajectory determination described above.

The vehicle 1400 may also include a power component 1424 configured to perform power management of the device 1400, a wired or wireless network interface 1426 configured to connect the device 1400 to a network, and an input output (I/O) interface 1428. The vehicle 1400 may operate based on an operating system stored in memory 1422, such as Window14 14erverTM,Mac O14 XTM,UnixTM,LinuxTM,FreeB14DTM or the like.

In an exemplary embodiment, a storage medium is also provided that includes instructions, such as memory 1422 including instructions, that are executable by a processor of the vehicle 1400 to perform the above-described method. The storage medium may be a non-transitory computer readable storage medium, which may be, for example, ROM, random Access Memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, etc.

In an exemplary embodiment, a computer program product is also provided, which, when being executed by a processor, may implement the above-mentioned method. The computer program product includes one or more computer instructions. When loaded and executed on a computer, these computer instructions may implement some or all of the methods described above, in whole or in part, in accordance with the processes or functions described in embodiments of the present disclosure.

Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in embodiments provided by the present disclosure may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, or the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory. By way of illustration, and not limitation, RAM can be in the form of a variety of forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), and the like.

The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.

The above examples merely represent a few implementations of the disclosed examples, which are described in more detail and are not to be construed as limiting the scope of the invention. It should be noted that it would be apparent to those skilled in the art that various modifications and improvements could be made to the disclosed embodiments without departing from the spirit of the disclosed embodiments. Accordingly, the protection scope of the disclosed embodiment patent should be subject to the appended claims.

Claims (12)

1. A method of security redundancy, the method comprising:

under a preset positioning failure scene, determining static obstacle information based on point cloud data acquired by a laser radar and a preset algorithm;

acquiring front vehicle track information sent by a front vehicle;

and determining the track information of the vehicle based on the static obstacle information, the track information of the front vehicle and the pre-acquired dynamic obstacle information.

2. The method of claim 1, wherein the static obstacle information includes road edge information and other obstacle information, and wherein the determining the static obstacle information based on the point cloud data collected by the lidar and a preset algorithm includes:

Determining a plurality of barrier lattices and attribute information of each barrier lattice according to the point cloud data and the preset algorithm;

and respectively determining the road edge information and the other barrier information according to the attribute information of the barrier lattices.

3. The method according to claim 2, wherein the determining the road edge information and the other obstacle information, respectively, according to the attribute information of the obstacle lattice, includes:

performing curve fitting processing according to the attribute information of the obstacle lattices, and determining the road edge information;

and updating according to the attribute information of the obstacle lattices to determine the other obstacle information.

4. The method according to claim 3, wherein the updating process based on the attribute information of the obstacle lattice, and determining the other obstacle information, comprises:

acquiring multi-frame smooth vehicle attitude information;

and updating the attribute information of each obstacle grid by using a preset probability model and a plurality of frames of the smooth vehicle posture information to obtain the other obstacle information.

5. The method of claim 3, wherein the determining the road edge information by performing curve fitting processing according to the attribute information of the obstacle lattice includes:

Performing curve fitting processing according to the attribute information of the obstacle lattices to obtain candidate curves;

and optimizing the candidate curves to obtain the road edge information.

6. The method according to claim 2, wherein the determining a plurality of obstacle lattices and attribute information of each of the obstacle lattices according to the point cloud data and the preset algorithm includes:

determining a plurality of candidate lattices and attribute information of each candidate lattice according to the point cloud data;

and updating the attribute information of each candidate lattice according to the sensing data of other sensors, and labeling categories to obtain the plurality of obstacle lattices and the attribute information of each obstacle lattice.

7. The method of claim 6, wherein determining a plurality of candidate lattices and attribute information for each of the candidate lattices from the point cloud data comprises:

determining a point cloud set forming a static obstacle according to the point cloud data;

and determining a plurality of candidate lattices and attribute information of each candidate lattice according to the point cloud set.

8. The method according to claim 1, wherein the method further comprises:

Post-processing the static obstacle information; the post-processing includes at least one of a clustering process, an attribute judging process, an elongation process, and a filtering process.

9. A safety redundant apparatus, the apparatus comprising:

the obstacle information determining module is used for determining static obstacle information based on point cloud data acquired by the laser radar and a preset algorithm under a preset positioning failure scene;

the acquisition module is used for acquiring the front vehicle track information sent by the front vehicle;

and the track information determining module is used for determining the track information of the vehicle based on the static obstacle information, the track information of the front vehicle and the pre-acquired dynamic obstacle information.

10. A vehicle comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the method of any one of claims 1 to 8 when the computer program is executed.

11. A storage medium having stored thereon a computer program, which when executed by a processor, implements the steps of the method of any of claims 1 to 8.

12. A computer program product comprising a computer program, characterized in that the computer program, when being executed by a processor, implements the steps of the method according to any one of claims 1-8.