JP7120565B2 - In-vehicle control device - Google Patents

Description

An embodiment of the present invention relates to an in-vehicle control device.

In conventional technology, when controlling the behavior of vehicles such as automobiles, machine learning represented by deep neural networks is used to autonomously control power, steering devices, and braking devices without human instructions or operations. By using a calculation module that uses , it realizes the action decision function that replaces human instructions and operations. This computing module has the function of determining the output signal from the input signal according to the internal conversion parameters. By using the output signal of , it is possible to determine the control signals for the power, steering system, and braking system corresponding to the action to be taken by the vehicle at that time. (For example, see Patent Document 1.)

JP 2017-211913 A

However, the conventional techniques described above do not consider the scale and complexity of the machine learning model, which are directly related to the difficulty of appropriately determining the transformation parameters inside the computing module. In general, as the size and complexity of machine learning models increase, machine learning needs to increase the amount of training data required for sufficient learning, increase the time required to complete learning, and partially However, there may be a problem such as being optimized only for Determining the control signals for the power, steering, and braking systems that correspond to the actions that the vehicle should take in various situations that may occur when driving a vehicle is a large-scale and complicated machine learning model. It is. For this reason, in the conventional technology, as described above as general problems of machine learning, the amount of learning data required for sufficient learning increases, the time required to complete learning increases, However, there is a problem that it can only be optimized in a specific way. In other words, the above-described conventional technique has a problem that it is very difficult to learn the arithmetic module for controlling the behavior of the vehicle.

SUMMARY OF THE INVENTION An object of the present invention is to provide an in-vehicle controller capable of reducing the difficulty of learning an arithmetic module for controlling the behavior of a vehicle.

According to an embodiment of the present invention, a plurality of hierarchized arithmetic modules transmit control information to be given to an in-vehicle device that controls the behavior of the vehicle based on detected information including at least information indicating that the situation around the vehicle is detected. and an acquisition unit for acquiring the detection information, and among the operation modules, based on the detection information acquired by the acquisition unit, the vehicle is selected from among a plurality of operation modules in the next hierarchy. A selection module that selects a computation module that performs computation according to surrounding conditions by computation using a neural network; and a control module that is selected from among the computation modules by the selection module and generates the control information based on the detection information. and an output unit that outputs the control information generated by the control module to the in-vehicle device, wherein the control modules are hierarchized, and the hierarchized control modules include: a pre-control module for determining a target behavior including at least a speed and a route of a vehicle based on said pre-control module; a post- stage control module that generates the control information to be provided to the in-vehicle device as speed control information and route control information to be implemented, and the plurality of arithmetic modules in the next hierarchy each have a different specific The selection module is an in-vehicle control device that selects, from among a plurality of computing modules, the computing module that has learned about the driving scene indicated by the detection information.

Further, in the in-vehicle control device of one embodiment of the present invention, the selection modules are hierarchically arranged, and the hierarchically arranged selection modules include conditions around the vehicle from among the plurality of selection modules in the next hierarchy. and a post-selection module that selects an arithmetic module from among a plurality of arithmetic modules in the next hierarchy, selected by the pre-selection module, according to the circumstances surrounding the vehicle. included.
Further, the selection modules are hierarchized, and the hierarchized selection modules include a pre-stage selection module that selects an operation module according to a situation around the vehicle from among a plurality of operation modules in the next hierarchy; a post-selection module that is selected by the pre-selection module and selects, from among a plurality of next-hierarchical processing modules, a computation module in accordance with a situation around the vehicle, wherein the pre-selection module Regarding the selection based on the situation, the self module selects the control module based on the correct rate of selection by the own module and the correct rate of selection by the latter selection module, or the control module is assigned to the latter selection module. Determine whether to select.

ADVANTAGE OF THE INVENTION According to this invention, the vehicle-mounted control apparatus which can reduce the difficulty of learning of the arithmetic module for controlling a behavior of a vehicle can be provided.

It is a figure showing an example of composition of a system containing an in-vehicle control device of a 1st embodiment. It is a figure which shows an example of the control information of this embodiment. It is a figure which shows an example of the calculation route of the vehicle-mounted control apparatus of this embodiment. It is a figure which shows an example of the flow of a process of the vehicle-mounted control apparatus of this embodiment. It is a figure which shows the 2nd example of the flow of a process of the vehicle-mounted control apparatus of this embodiment. It is a figure which shows the 3rd example of the flow of a process of the vehicle-mounted control apparatus of this embodiment. 3 is a diagram showing a specific example of the configuration of an arithmetic module according to the embodiment; FIG. It is a figure which shows an example of the learning procedure of the arithmetic module of this embodiment. It is a figure which shows an example of the flow of a process of the vehicle-mounted control apparatus of 2nd Embodiment. It is a figure which shows an example of the flow of learning of the arithmetic module provided to the vehicle-mounted control apparatus of this embodiment.

[First embodiment]
An overview of the in-vehicle control device 10 of the present embodiment will be described below with reference to the drawings.
FIG. 1 is a diagram showing an example of the configuration of a system including an in-vehicle control device 10 of this embodiment.

The vehicle CR includes an in-vehicle control device 10 , a detection information generation unit 20 , an in-vehicle device 30 , a logger 40 and a communication unit 50 . The in-vehicle control device 10 and the detection information generation unit 20, and the in-vehicle control device 10 and the in-vehicle device 30 are connected to each other by an in-vehicle communication network such as CAN (Controller Area Network). It is possible to give and receive.
In the present embodiment, the in-vehicle control device 10 is described as a device having a housing independent from the detection information generation unit 20 and the in-vehicle device 30, but the present invention is not limited to this. The in-vehicle control device 10 may be configured as a device integrated with the detection information generating section 20 and the in-vehicle device 30 .

The detection information generation unit 20 detects the operation of each part of the vehicle CR and the surrounding conditions of the vehicle CR, and generates detection information. The detection information generated by the detection information generation unit 20 is also referred to as detection result DR. The detection result DR may include the detection result of the operation of each part of the vehicle CR and the detection result of the situation around the vehicle CR, but the detection result of the situation around the vehicle CR will be described below. , the description of the detection result of the operation of each part of the vehicle CR is omitted.

The detection information generation section 20 includes a sensor section 200 and a detection result processing section 210 .
The sensor unit 200 is a sensor or the like that detects the surrounding conditions of the vehicle CR, and outputs the detection result to the in-vehicle control device 10 via the detection result processing unit 210 or directly as the detection result DR. As an example of this embodiment, the sensor unit 200 includes a LIDAR device 201 , a millimeter wave device 202 , a GNSS device 203 and a camera device 204 .
The LIDAR device 201 performs LIDAR (Light Detection and Ranging, Laser Imaging Detection and Ranging) image acquisition and ranging around the vehicle CR. The LIDAR device 201 outputs the image acquisition result and the distance measurement result as the detection result DR.
The millimeter wave device 202 performs shape detection and distance measurement around the vehicle CR using millimeter waves. The millimeter wave device 202 outputs the shape detection result and the distance measurement result as the detection result DR.
The GNSS device 203 receives positioning signals from positioning satellites by GNSS (Global Navigation Satellite System). The GNSS device 203 outputs the received positioning signal as the detection result DR.
The camera device 204 uses visible light and infrared light to obtain an image of the surroundings of the vehicle CR and perform distance measurement. The camera device 204 outputs the image acquisition result and the distance measurement result as the detection result DR.

The detection result processing unit 210 processes the detection result DR output by the sensor unit 200 to generate information indicating the circumstances around the vehicle CR. In one example of this embodiment, the detection result processing unit 210 includes an image recognition unit 211 , an object detection unit 212 and a position estimation unit 213 .
Based on the image acquisition results and distance measurement results output by the LIDAR device 201 and camera device 204, the image recognition unit 211 recognizes the presence or absence of an object included in these images, the shape, size, and type of the object, and the distance to the object. , the speed of movement of an object, etc. The image recognition unit 211 outputs the image recognition result to the in-vehicle control device 10 as the detection result DR.
The object detection unit 212 detects the presence or absence of an object, the shape/size/type of the object, the distance to the object, the moving speed of the object, etc., based on the shape detection result and the distance measurement result output by the millimeter wave device 202. . The object detection unit 212 outputs the object detection result to the vehicle-mounted controller 10 as the detection result DR.
The position estimator 213 estimates the position of the vehicle based on the positioning signal output by the GNSS device 203 . The position estimation unit 213 outputs the position estimation result of the own vehicle to the in-vehicle control device 10 as the detection result DR.

The in-vehicle device 30 includes an ECU (Electronic Control Unit) 300 and actuators (not shown), and controls the behavior of the vehicle CR. The ECU 300 includes, for example, a travel control ECU 301 that drives the engine and a travel motor, a brake control ECU 302 that controls the brake, a steering ECU 303 that controls the steering angle, and a lamp that controls lights such as headlights and emergency stop lights. There is a control ECU 304 and the like.
These ECUs 300 operate based on control information CMD output from the in-vehicle control device 10, and output information and signals for driving actuators.

The in-vehicle control device 10 includes a CPU (Central Processing Unit) (not shown), performs calculation based on the detection result DR generated by the detection information generation unit 20, and generates control information CMD as a result of this calculation. . The in-vehicle control device 10 of the present embodiment performs this calculation using a plurality of hierarchized calculation modules MDL.

That is, the in-vehicle control device 10 provides control information CMD to the in-vehicle device 30 that controls the behavior of the vehicle CR based on the detection result (detection information) DR including at least the information that the situation around the vehicle CR is detected. It is generated by a plurality of hierarchized arithmetic modules MDL.
The in-vehicle control device 10 includes an acquisition unit 120 (not shown), an arithmetic module MDL, and an output unit 130 (not shown) as functional units implemented by software or hardware.

The acquisition unit 120 acquires the detection result DR output by the detection information generation unit 20 . The acquisition unit 120 normalizes the acquired detection result DR and supplies the normalized information to the computation module MDL. For example, when the LIDAR device 201 is a device that detects 16 orientations, the acquisition unit 120 sets the sensor value of the first orientation, the sensor value of the second orientation, . Normalize to values in the range .0. The acquisition unit 120 supplies the normalized values to the computation module MDL as a data string in which the values are arranged in the order of the first direction, the second direction, . . . , the 16th direction.

The output unit 130 outputs the control information CMD generated by the arithmetic module MDL to the in-vehicle device 30 . An example of the control information CMD output by the output unit 130 will be described with reference to FIG.

FIG. 2 is a diagram showing an example of control information CMD of this embodiment. Operations that are possible for the vehicle CR are defined in advance as control information CMD. The output unit 130 outputs a value of "1" when operating each of these control information CMDs, and a value of "0 (zero)" when not operating. In the example shown in the figure, the output unit 130 outputs the value "1" for each of the control information CMD "turn the steering wheel to the right" and the control information CMD "depress the accelerator". As a result of outputting this control information CMD, the ECU 300 of the in-vehicle device 30 controls the steering angle and the power for running, so that the vehicle CR "accelerates while turning to the right".

Returning to FIG. 1, the computing module MDL has a selection module 100 and a control module 110 .
Based on the detection result DR acquired by the acquisition unit 120, the selection module 100 selects a computation module MDL that performs computation according to the circumstances around the vehicle CR from among the plurality of computation modules MDL in the next hierarchy. This selection module 100 is operated by a neural network NN (eg deep neural network).

The control module 110 is selected by the selection module 100 and generates control information CMD based on the detection result DR.
That is, the computing modules MDL are hierarchized with the selection module 100 as the preceding computing module MDL and the control module 110 as the succeeding computing module MDL.

[Hierarchy of control modules]
Hierarchization of the control module 110 will be described. The hierarchized control modules 110 include a front-stage control module 110a and a rear-stage control module 110b. Among them, the pre-stage control module 110a determines the target behavior TB of the vehicle CR based on the detection result DR. The post-control module 110b generates control information CMD to be provided to the in-vehicle device 30 based on the target behavior TB of the vehicle CR determined by the pre-control module 110a.

Here, the target behavior TB of the vehicle CR is "acceleration while turning right" in the example shown in FIG. In this example, the vehicle CR is waiting to turn right in the right turn lane of the intersection, and turns right at the intersection following the progress of the preceding vehicle that is also waiting to turn right. The detection information generation unit 20 sends the detection result DR indicating that the vehicle CR is in the right turn lane of the intersection and the detection result DR indicating that the vehicle in front of the vehicle CR has advanced and started to turn right to the in-vehicle control device 10. supply. Based on these detection results DR, the pre-stage control module 110a determines that the target behavior TB of the vehicle CR is "accelerate while turning right". Based on the target behavior TB of the vehicle CR determined by the preceding control module 110a, i.e., "accelerate while turning right", the post-stage control module 110b sets "turn the steering wheel to the right: value 1" and control information CMD "Step on accelerator: value 1" is generated as control information CMD.
In the example shown in FIG. 1, among the control modules 110, the control modules 111 and 112 correspond to the front-stage control module 110a, and the control modules 115 and 116 correspond to the rear-stage control module 110b. That is, in one example of the same figure, the control module 111 and the control module 115 are hierarchized with each other, and the control module 112 and the control module 116 are hierarchized with each other.

[Hierarchy of selected modules]
Layering of the selection module 100 will be described. The hierarchical selection modules 100 include a front selection module 100a and a rear selection module 100b. Among them, the preceding selection module 100a selects the selection module 100 according to the circumstances around the vehicle CR from among the plurality of selection modules 100 in the next hierarchy. The post-stage selection module 100b is selected by the pre-stage selection module 100a, and selects a computation module MDL according to the circumstances around the vehicle CR from among a plurality of computation modules MDL in the next hierarchy.

In the example shown in FIG. 1, in the relationship between the selection module 101 and the selection module 102 among the selection modules 100, the selection module 101 corresponds to the preceding selection module 100a, and the selection module 102 corresponds to the subsequent selection module 100b. . In the relationship between the selection module 102 and the selection module 103, the selection module 102 corresponds to the preceding selection module 100a, and the selection module 103 corresponds to the subsequent selection module 100b.

The logger 40 is connected to the in-vehicle control device 10, the detection information generating section 20, and the in-vehicle device 30, and temporarily stores information output by each of these units and devices.
The communication unit 50 transmits and receives information to and from the data center DC via the base station BS. Specifically, the communication unit 50 transmits information temporarily stored in the logger 40 to the data center DC. In addition, the communication unit 50 receives update data for the computing module MDL provided by the data center DC.

The data center DC re-learns the computing module MDL based on the log data provided from the vehicle CR to update the computing module MDL, and distributes the updated computing module MDL to the vehicle CR. Specifically, the data center DC includes a log database DC1, a relearning section DC2, and a module library DC3. Log data transmitted from the vehicle CR is accumulated in the log database DC1. This log data includes the detection result DR, control information CMD, target behavior TB, and behavior of the vehicle CR. The relearning unit DC2 updates the computing module MDL by re-learning the computing module MDL based on the log data accumulated in the log database DC1. The module library DC3 stores the arithmetic module MDL updated by the relearning unit DC2. The updated computation module MDL stored in the module library DC3 is provided to the vehicle CR in response to a request from the vehicle CR, a request from the manager of the computation module MDL, or the like.

Next, with reference to FIG. 3, the calculation path of the in-vehicle control device 10 will be described.
FIG. 3 is a diagram showing an example of a calculation path of the in-vehicle control device 10 of this embodiment. The selection module 101 acquires the detection result DR from the acquisition unit 120 (not shown). The selection module 101 determines which of the selection module 102 and the control module 112 should be selected based on the detection result DR. In this example, selection module 101 selects selection module 102 . Next, the selection module 102 determines which of the selection module 103 and the control module 111 should be selected based on the detection result DR. In this example, selection module 102 selects control module 111 . The control module 111 determines a target behavior TB of the vehicle CR and supplies the target behavior TB to the control module 115 . The control module 115 generates control information CMD based on the target behavior TB. The generated control information CMD is supplied to each ECU 300 by the output unit 130 (not shown).

[Processing flow of in-vehicle control device (embodiment 1)]
FIG. 4 is a diagram showing an example of the processing flow of the in-vehicle control device 10 of this embodiment. In this example, the in-vehicle control device 10 includes one selection module 100 and two control modules 110 as computation modules MDL. In this example, the arithmetic module MDL of the vehicle-mounted control device 10 is hierarchized into two layers, a selection module 100 in the first stage and a control module 110 in the second stage.

(Step S10) The sensor unit 200 detects the surrounding conditions of the vehicle CR and outputs the detected result as the detection result DR.
(Step S20) Based on the detection result DR output in step S10, the detection result processing unit 210 detects the presence or absence of the object, the shape/size/type of the object, the distance to the object, the moving speed of the object, and the like. . The detection result processing unit 210 outputs the detection result to the in-vehicle control device 10 as the detection result DR.
(Step S<b>30 ) The acquisition unit 120 of the in-vehicle control device 10 acquires the detection result DR output from the sensor unit 200 or the detection result DR output from the detection result processing unit 210 .

(Step S40) The selection module 100 selects the control module 110 from among the plurality of control modules 110 according to the circumstances around the vehicle CR indicated by the detection result DR acquired in step S30.
(Step S50) The control module 110 generates control information CMD based on the detection result DR. If the selection module 100 selects the control module 110-A (step S40; A), the control module 110-A executes step S50-1. If the selection module 100 selects the control module 110-B (step S40; B), the control module 110-B executes step S50-2.

(Step S60) Each ECU 300 controls the behavior of the vehicle CR by driving actuators based on the control information CMD generated in step S50. For example, the traveling control ECU 301 drives the engine and the traveling motor based on the control information CMD (step S60-1). The brake control ECU 302 controls the brake based on the control information CMD (step S60-2). The steering ECU 303 controls the steering angle based on the control information CMD (step S60-3).

[Processing flow of in-vehicle control device (embodiment 2)]
Next, a second example of the processing flow of the in-vehicle control device 10 will be described with reference to FIG.
FIG. 5 is a diagram showing a second example of the processing flow of the in-vehicle control device 10 of this embodiment. In the following description, the same reference numerals are given to the same parts and operations as those described in the example of FIG. 4, and the description thereof will be omitted. This embodiment differs from the first embodiment shown in FIG. 4 in that the control module 110 is divided into a plurality of hierarchies.

(Step S150) The front stage control module 110a generates a target behavior TB of the vehicle CR based on the detection result DR. If the selection module 100 selects the preceding control module 110a-A (step S40; A), the preceding control module 110a-A executes step S150-1. If the selection module 100 selects the preceding control module 110a-B (step S40; B), the preceding control module 110a-B executes step S150-2.

(Step S160) The post-control module 110b generates control information CMD based on the target behavior TB of the vehicle CR and the detection result DR. If the selection module 100 selects the post-control module 110b-A (step S150; A), the post-control module 110b-A executes step S160-1. If the selection module 100 selects the post-control module 110b-B (step S150; B), step S160-2 is executed by the post-control module 110b-B.

[Processing flow of in-vehicle control device (embodiment 3)]
Next, a third example of the processing flow of the in-vehicle control device 10 will be described with reference to FIG.
FIG. 6 is a diagram showing a third example of the processing flow of the in-vehicle control device 10 of this embodiment. In the following description, the same reference numerals are given to the same units and operations as those described in the examples of FIGS. 4 and 5, and the description thereof will be omitted. This embodiment differs from the first embodiment shown in FIG. 4 in that the selection module 100 is divided into a plurality of layers.

(Step S240) The front-stage selection module 100a selects from among the plurality of rear-stage selection modules 100b (the rear-stage selection modules 100b-A and 100b-B) the situation around the vehicle CR indicated by the detection result DR obtained in step S30. The appropriate post-selection module 100b is selected.
(Step S250) The latter selection module 100b selects the vehicle indicated by the detection result DR obtained in step S30 from among the plurality of control modules 110 (control modules 110-A1, 110-A2, 110-B1 and 110-B2). The control module 110 is selected according to the circumstances surrounding the CR.

Note that the computation module MDL may be configured by combining the above-described second and third embodiments.

[Concrete Example of Operation Module Configuration of In-vehicle Control Device]
Next, a specific example of the configuration of the arithmetic module MDL of this embodiment will be described with reference to FIG.
FIG. 7 is a diagram showing a specific example of the configuration of the computation module MDL of this embodiment. The selection module 100 is composed of four layers of selection modules 101 to 104 . Each of the selection modules 101 to 104 performs calculations using a neural network NN.

The control module 110 is classified into three control modes MD. This control mode MD includes a yield mode MD1, a normal running mode MD2, and an emergency action mode MD3.
Yield mode MD1 is composed of acceleration and deceleration at a roundabout (control module 111) and lane running (control module 115).
The normal driving mode MD2 includes a roundabout/intersection vicinity driving mode MD2-1 and a straight road driving mode MD2-2. The roundabout/intersection periphery driving mode MD2-1 further includes an acceleration/deceleration mode MD2-1-1 and a stop mode MD2-1-2.
The acceleration/deceleration mode MD2-1-1 is composed of acceleration/deceleration (control module 113a) corresponding to the oncoming vehicle/merging vehicle and lane travel (control module 117a).
The stop mode MD2-1-2 is composed of stopping at a temporary stop line (control module 113b) and lane running (control module 117b).
The straight road running mode MD2-2 is composed of acceleration and deceleration (control module 113) corresponding to the preceding vehicle and lane running (control module 117).
The emergency action mode MD3 is composed of emergency stop (control module 112) and lane running (control module 116).

Here, the acceleration/deceleration at the roundabout (control module 111) and the emergency stop (control module 112) are both calculated by the neural network NN.
Lane running (control module 115, control module 116, control module 117, control module 117a, and control module 117b) all perform calculations using a neural network NN and a PID (Proportional-Integral-Differential) control algorithm.
Acceleration/deceleration according to the preceding vehicle (control module 113), acceleration/deceleration according to the oncoming vehicle/merging vehicle (control module 113a), and stopping at the stop line (control module 113b) are all performed by rule-based calculations. perform calculations. The rule-based calculation is a calculation procedure for calculating the control information CMD to be given to the in-vehicle device 30, in which the correspondence relationship between the detection result DR and the control information CMD is determined in advance. An example of rule-based computation is a computation method that calculates control information CMD by conditional branching by if-then-else.

The selection module 101 selects either the selection module 102 or the control module 110 in the emergency action mode MD3 based on the detection result DR acquired by the acquisition unit 120 . As an example, when an emergency situation such as a person jumping out of a side road or a load falling from the vehicle ahead occurs while the vehicle CR is running, the selection module 101 selects the control module 110 in the emergency action mode MD3. do. Selection module 101 selects selection module 102 if an emergency situation such as this example does not occur.

The selection module 102 selects either the yield mode MD1 or the selection module 103 based on the detection result DR. As an example, in a state in which a vehicle CR is traveling on a roundabout (roundabout), there is a case where the entire vehicle group can travel efficiently by yielding to each other and traveling. In such a case, selection module 102 selects yield mode MD1. Selection module 102 selects selection module 103 if the roundabout is not being run.

The selection module 103 selects either the selection module 104 or the straight road running mode MD2-2. As an example, the selection module 103 selects the selection module 104 when the vehicle CR runs through a roundabout or an intersection. When the vehicle CR runs on a straight road, the selection module 103 selects the straight road running mode MD2-2.

The selection module 104 selects either the acceleration/deceleration mode MD2-1-1 or the stop mode MD2-1-2. As an example, when the vehicle CR is presenting a stop (red) signal at an intersection with a traffic light or is traveling through an intersection with a stop sign, the selection module 104 selects the stop mode MD2-1-2. select. The selection module 104 selects the acceleration/deceleration mode MD2-1-1 when the vehicle CR merges into the roundabout or follows the preceding vehicle in the roundabout.

When selected by the selection module 100, the control module 110 belonging to each control mode MD outputs the calculation result to each ECU 300 via the output unit 130 as control information CMD.

[Learning Procedure of Operation Module MDL]
FIG. 8 is a diagram showing an example of the learning procedure of the computation module MDL of this embodiment. As an example, the control module 110 is configured in two layers of a front-stage control module 110a and a rear-stage control module 110b, and the selection module 100 selects a specific front-stage control module 110a from among a plurality of front-stage control modules 110a. will be described.

In the first stage, learning (for example, machine learning) of the post-stage control module 110b is performed (FIG. 8A). This post-stage control module 110b includes a PID control algorithm. In the first stage, learning is made as to what values should be set for the parameters of this PID control.
In the second stage, the learned rear-end control module 110b is connected to the front-end control module 110a, and learning of the front-end control module 110a is performed (FIG. 8(B)). In the second stage, the system designer determines the driving scene for which the front-end control module 110a is to be used for each front-end control module 110a, and learns the front-end control module 110a based on the driving scene data. That is, the pre-stage control module 110a is made to learn only a specific driving scene.
In the third stage, the learned rear-stage control module 110b and the front-stage control module 110a are connected to the selection module 100, and the selection module 100 is made to learn (FIG. 8(C)). In the third stage, while observing the behavior of the pre-control module 110a with respect to the detection result DR, it is searched which pre-control module 110a should be selected for a certain detection result DR. As a result, in the third stage, the selection module 100 is made to learn what driving scene a given detection result DR indicates.

That is, in this example, there are two or more computing modules MDL in the latter stage of a certain computing module MDL. It is connected to the arithmetic module MDL in the preceding stage. The former module judges the situation from the detection result DR of the sensor unit 200 and the detection result processing unit 210 obtained while the automobile is running, and selects one of the latter operation modules MDL according to the situation at that time. to operate. The operating module MDL in the subsequent stage determines the control information CMD of the power, steering device, and braking device from the detection result DR of the sensor unit 200 and the detection result processing unit 210 .

[Summary of the first embodiment]
By configuring the computing modules MDL as described above, the in-vehicle control device 10 can make each computing module MDL learn, and can make other computing modules MDL learn using the learned computing module MDL. can be done. Therefore, according to the in-vehicle control device 10, the learning efficiency is higher than when all the arithmetic modules MDL are learned at the same time, and the learning can be performed in a short period of time. That is, according to the in-vehicle control device 10, the learning difficulty of the arithmetic module MDL for controlling the behavior of the vehicle CR can be reduced.

Further, as described above, the arithmetic module MDL of the in-vehicle control device 10 is configured by dividing the control module 110 into a plurality of hierarchies. That is, the in-vehicle control device 10 includes a pre-control module 110a that determines the target behavior TB of the vehicle CR (for example, a travel plan that indicates which point should be passed at what speed after how many seconds), and a given target and a post-control module 110b that determines control signals for the power, steering and braking systems so that the behavior TB (eg, the trip plan described above) can be achieved. Compared to the configuration of the first embodiment described above, although the total number of operation modules MDL that require machine learning is increased, the abstraction level of the problem to be solved by each operation module MDL is lower than before division. Become. Therefore, the scale and complexity of the machine learning model applied to each control module 110 are reduced. That is, according to the in-vehicle control device 10, it is possible to reduce the learning difficulty of the arithmetic module MDL (control module 110) for controlling the behavior of the vehicle CR.

Further, as described above, the calculation module MDL of the in-vehicle control device 10 is configured by dividing the selection module 100 into a plurality of hierarchies. That is, in the vehicle-mounted control device 10, the selection module 100 is hierarchized like the pre-selection module 100a and the post-selection module 100b, and is composed of a plurality of modules arranged in a tree structure. The selection module 100 at each stage of this tree structure selects and operates one of the operation modules MDL connected behind itself, thereby progressing situation determination step by step. Compared to the configuration of the first embodiment described above, the total number of modules that need to be machine-learned has increased, but the abstraction level of the problem to be solved by the selection module 100 has become lower than before division. For example, the size and complexity of the machine learning model applied to the selection module 100 is distributed throughout the tree structure of the selection module 100 . Therefore, the scale and complexity of the machine learning model applied to each selection module 100 are reduced. That is, according to the in-vehicle control device 10, it is possible to reduce the learning difficulty of the arithmetic module MDL (selection module 100) for controlling the behavior of the vehicle CR.

[Second embodiment]
An in-vehicle control device 10 according to the second embodiment will be described below. In the following description, the same reference numerals are given to the same units and operations as those described in the first embodiment, and the description thereof will be omitted. This embodiment differs from the first embodiment in that it includes a selection module 100 that selects according to a rule-based operation. Further, the present embodiment differs from the first embodiment in that the selection module 100 selects the next-stage operation module MDL based on the accuracy rate of the selection candidate operation module MDL.

[Selection of calculation module by rule-based calculation]
The selection module 100 of this embodiment includes a first selection module 100NN and a second selection module 100RB.
Based on the detection result (detection information) DR acquired by the acquisition unit 120, the first selection module 100NN is a computation module that performs computation according to the surrounding conditions of the vehicle CR from among a plurality of computation modules MDL in the next hierarchy. MDL is selected by computation by neural network NN.
The second selection module 100RB is a calculation module that performs calculation according to the surrounding conditions of the vehicle CR from among a plurality of calculation modules MDL in the next hierarchy based on the detection result (detection information) DR obtained by the obtaining unit 120. The MDL is selected by rule-based calculation in which the correspondence between the detection result (detection information) DR and the control information CMD is predetermined.
The control module 110 is selected by either the first selection module 100NN or the second selection module 100RB described above, and generates control information CMD based on the detection result (detection information) DR.

The control module 110 also includes a rule-based control module 110RB. This rule-based control module 110RB generates control information CMD to be given to the in-vehicle device 30 by rule-based calculation based on the detection result DR.

[Selection of calculation module based on accuracy rate]
The selection module 100 is hierarchical. The hierarchical selection modules 100 include a front selection module 100a and a rear selection module 100b. Of these, the front selection module 100a is controlled by the front selection module 100a based on the selection accuracy rate VLD of the self module and the selection accuracy rate VLD of the rear selection module 100b for selection based on the circumstances surrounding the vehicle CR. It is determined whether the module 110 is selected or the control module 110 is selected by the subsequent selection module 100b. Here, the accuracy rate VLD refers to the degree of correctness of the information output by the computation module MDL. For example, the accuracy rate VLD is indicated by the percentage of correct results selected by the selection module 100 according to the circumstances surrounding the vehicle CR. Further, for example, the accuracy rate VLD is indicated by the rate at which the control information CMD that makes the behavior of the vehicle CR correct (that is, the correct control information CMD) can be output among the control information CMD output by the control module 110 . In these cases, the "correct selection result" and the "correct behavior" of the vehicle CR according to the circumstances around the vehicle CR are defined in advance for each situation around the vehicle CR.

[Processing flow for selection of computing module]
Next, an example of the processing flow of the in-vehicle control device 10 will be described with reference to FIG. 9 .
FIG. 9 is a diagram showing an example of the processing flow of the in-vehicle control device 10 of this embodiment. In this example, the in-vehicle control device 10 includes four selection modules 100, selection modules A to D, and eight control modules 110, control modules A1 to D2.
The selection module A is the pre-selection module 100a by the neural network NN, ie the first selection module 100NN. The selection module B is a pre-selection module 100a based on rule-based operations, that is, a second selection module 100RB. Selection module C and selection module D are post-selection modules 100b.
Each of the control modules A1-D2 generates control information CMD according to the circumstances around a certain vehicle CR.

(Step S310) The selection module A selects the selection module 100 with the higher accuracy rate VLD from the selection module A or the selection module C. Here, the selection module A estimates the accuracy rate VLD when each of the selection module A and the selection module C is selected. The selection module A selects the selection module 100 with the highest estimated accuracy rate VLD from among the selection modules 100 . Further, when the accuracy rate VLD of both the selection module A and the selection module C does not satisfy a certain reference value, the selection module A selects the selection module B by computation of the neural network NN.
(Step S320) When the selection module A is selected in step S310, the selection module A selects the control module 110 suitable for the circumstances around the vehicle CR from among the control modules A1 and A2.
(Step S330) When the selection module B is selected in step S310, the selection module B determines whether or not there is a rule corresponding to the detection result DR. If it is determined that there is a rule (step S330; YES), the selection module B selects the control module 110 suitable for the circumstances around the vehicle CR from the control module B1 or the control module B2 by rule-based calculation. Select (step S340).
(Step S350) If the selection module C is selected in step S310, the selection module C selects the control module 110 suitable for the circumstances around the vehicle CR, from the control module C1 or the control module C2.
(Step S360) If it is determined that there is no rule in step S330 (step S330; NO), the selection module D selects the control module D1 or D2 that is suitable for the circumstances around the vehicle CR. Select 110.

(Steps S370-1 to -8) When selected by the selection module 100, the control modules A1 to D2 generate control information CMD based on the detection result DR.

[Flow of learning of computing module]
Next, an example of the learning flow of the computation module MDL will be described with reference to FIG.
FIG. 10 is a diagram showing an example of the learning flow of the arithmetic module MDL provided in the vehicle-mounted control device 10 of the present embodiment. Here, the arithmetic module MDL is provided to the vehicle-mounted control device 10 after being learned by the re-learning unit DC2 of the data center DC. The flow of learning of the arithmetic module MDL by the relearning unit DC2 will be described below. In this example, the computation module MDL is composed of one pre-selection module 100a, two post-selection modules 100b (100b-1 and 100b-2), and three control modules 110 (110-1 to -3). be.

(Step S400) The relearning unit DC2 acquires the detection result DR from the log database DC1.
(Step S410) The relearning unit DC2 supplies the acquired detection result DR to the pre-selection module 100a to estimate the accuracy rate VLD of the calculation module MDL, and select the calculation module MDL having the high accuracy rate VLD. let them learn.
(Step S420) The re-learning unit DC2 supplies the detection result DR to the pre-stage selection module 100a, and performs learning for classifying (determining) the circumstances around the vehicle CR from the detection result DR.
(Steps S430 and S440) The relearning unit DC2 supplies the detection result DR to the subsequent selection module 100b selected by the preceding selection module 100a and the control module 110, and causes them to generate control information CMD.
(Steps S450, S460) The relearning unit DC2 supplies the control information CMD generated in step S440 (steps S440-1 to -3) to the ECU 300 (or an emulator of the ECU 300). Observe the output and The re-learning unit DC2 learns the arithmetic module MDL using, as teacher data, the adequacy (good or bad) of the control result obtained by comparing the observed output of the ECU 300 and the desired (correct) behavior of the vehicle CR.

[Summary of the second embodiment]
By configuring the computing modules MDL as described above, the in-vehicle control device 10 can make each computing module MDL learn, and can make other computing modules MDL learn using the learned computing module MDL. can be done. Therefore, according to the in-vehicle control device 10, the learning efficiency is higher than when all the arithmetic modules MDL are learned at the same time, and the learning can be performed in a short period of time. That is, according to the in-vehicle control device 10, the learning difficulty of the arithmetic module MDL for controlling the behavior of the vehicle CR can be reduced.

The vehicle-mounted control device 10 of the present embodiment also includes a selection module 100 (second selection module 100RB) that selects the computation module MDL by rule-based computation. For example, there are situations where the vehicle CR is required to behave in a predetermined manner according to traffic regulations, such as a stop (red) signal, a caution (yellow) signal, and a stop sign. For such a predetermined behavior, if a rule-based operation such as if-then-else conditional branching is used, the number of situations to be learned by the neural network NN can be further reduced.
More specifically, for some of the situations that the vehicle CR encounters, it is necessary to define under what conditions the situation is defined and what behavior should be taken in that situation. It is clearly defined whether or not In this case, the in-vehicle control device 10 of the present embodiment determines the situation by the selection module 100 (second selection module 100RB) based on rule-based calculations based on predetermined conditions for situations in which the behavior is clearly defined. However, for situations where the behavior is not clearly defined, the situation is determined by the selection module 100 (first selection module 100NN) based on the neural network NN. With this configuration, the in-vehicle control device 10 can classify various situations while reducing the amount of learning data compared to the case where all the selection modules 100 are based on the neural network NN. can be done.
That is, the in-vehicle control device 10 of the present embodiment can reduce the learning difficulty of the arithmetic module MDL (selection module 100) for controlling the behavior of the vehicle CR.

The vehicle-mounted control device 10 of this embodiment also includes a rule-based control module 110RB that generates control information CMD by rule-based calculation. If a part of the control information CMD is generated by a rule-based operation, it is possible to reduce the number of situations to be learned by the neural network NN. That is, the in-vehicle control device 10 of the present embodiment can reduce the learning difficulty of the arithmetic module MDL (control module 110) for controlling the behavior of the vehicle CR.

Further, the in-vehicle control device 10 of the present embodiment determines which operation module MDL should be selected by the pre-stage selection module 100a based on the accuracy rate VLD. By selecting the computation module MDL based on the accuracy rate VLD in this way, the in-vehicle control device 10 can select the computation module MDL more suitable for the circumstances around the vehicle CR.

Although embodiments of the present invention and variations thereof have been described above, these embodiments and variations thereof are presented as examples and are not intended to limit the scope of the invention. These embodiments and modifications thereof can be implemented in various other forms, and various omissions, replacements, and modifications can be made without departing from the scope of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are included in the scope of the invention described in the claims and equivalents thereof.

Each of the devices described above has a computer inside. The process of each process of each device described above is stored in a computer-readable recording medium in the form of a program, and the above process is performed by reading and executing this program by a computer. Here, the computer-readable recording medium refers to magnetic disks, magneto-optical disks, CD-ROMs, DVD-ROMs, semiconductor memories, and the like. Alternatively, the computer program may be distributed to a computer via a communication line, and the computer receiving the distribution may execute the program.

Further, the program may be for realizing part of the functions described above.
Further, it may be a so-called difference file (difference program) that can realize the above-described functions in combination with a program already recorded in the computer system.

DESCRIPTION OF SYMBOLS 10... Vehicle-mounted control apparatus, 100... Selection module, 100a... Pre-stage selection module, 100b... Subsequent-stage selection module, 100NN... First selection module, 100RB... Second selection module, 110... Control module, 110RB... Rule base control module , 120... Acquisition unit, 130... Output unit CMD... Control information, DR... Detection result, MDL... Operation module, VLD... Accuracy rate

Claims (3)

An in-vehicle control device that generates control information to be given to an in-vehicle device that controls the behavior of the vehicle by using a plurality of hierarchized arithmetic modules, based on detected information including at least information about the situation surrounding the vehicle. hand,
an acquisition unit that acquires the detection information;
Based on the detection information acquired by the acquisition unit, among the computation modules, a computation module that performs computation according to the situation around the vehicle is selected from among a plurality of computation modules in the next hierarchy by computation using a neural network. a selection module;
a control module selected by the selection module among the operation modules and generating the control information based on the detection information;
an output unit that outputs the control information generated by the control module to the in-vehicle device;
with
The control modules are hierarchical,
The hierarchical control module includes:
a pre-control module that determines a target behavior including at least the speed and route of the vehicle based on the detected information;
Based on the speed and the route included in the target behavior determined by the pre-stage control module, the control information to be given to the in-vehicle device as speed control information and route control information for realizing the target behavior is generated. a post-stage control module that
includes
The plurality of computing modules in the next hierarchy are computing modules that have learned about specific driving scenes different from each other,
The in-vehicle control device, wherein the selection module selects, from among a plurality of calculation modules, a calculation module that has learned about the driving scene indicated by the detection information. The selection modules are hierarchical,
The hierarchical selection module includes:
a previous stage selection module that selects the selection module according to the circumstances around the vehicle from among the plurality of selection modules in the next hierarchy;
a post-selection module that is selected by the pre-selection module and selects, from among a plurality of next-hierarchical processing modules, a computation module that corresponds to a situation around the vehicle;
The in-vehicle control device according to claim 1 , comprising: The selection modules are hierarchical,
The hierarchical selection module includes:
a pre-stage selection module that selects a computation module according to the surrounding conditions of the vehicle from among a plurality of computation modules in the next hierarchy;
a post-selection module that is selected by the pre-selection module and selects, from among a plurality of next-hierarchical processing modules, a computation module that corresponds to a situation around the vehicle;
includes
The pre-selection module is
Based on the selection accuracy rate of the self module and the selection accuracy rate of the post-selection module regarding the selection based on the circumstances surrounding the vehicle, the self-module selects the control module or the post-selection module selects the control module. The in-vehicle control device according to claim 1 or 2 , wherein it is determined whether to select the control module.

Images