WO2017181319A1 - Particle swarm optimization and reinforcement learning algorithm-based dynamic walking control system for biomimetic biped robot - Google Patents
Particle swarm optimization and reinforcement learning algorithm-based dynamic walking control system for biomimetic biped robot Download PDFInfo
- Publication number
- WO2017181319A1 WO2017181319A1 PCT/CN2016/079543 CN2016079543W WO2017181319A1 WO 2017181319 A1 WO2017181319 A1 WO 2017181319A1 CN 2016079543 W CN2016079543 W CN 2016079543W WO 2017181319 A1 WO2017181319 A1 WO 2017181319A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- control center
- controller
- signal input
- signal output
- control
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Ceased
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B62—LAND VEHICLES FOR TRAVELLING OTHERWISE THAN ON RAILS
- B62D—MOTOR VEHICLES; TRAILERS
- B62D57/00—Vehicles characterised by having other propulsion or other ground- engaging means than wheels or endless track, alone or in addition to wheels or endless track
- B62D57/02—Vehicles characterised by having other propulsion or other ground- engaging means than wheels or endless track, alone or in addition to wheels or endless track with ground-engaging propulsion means, e.g. walking members
- B62D57/032—Vehicles characterised by having other propulsion or other ground- engaging means than wheels or endless track, alone or in addition to wheels or endless track with ground-engaging propulsion means, e.g. walking members with alternately or sequentially lifted supporting base and legs; with alternately or sequentially lifted feet or skid
Definitions
- the invention relates to a bionic biped robot dynamic walking control system based on particle swarm optimization and reinforcement learning algorithm, and belongs to the technical field of bionic humanoid robot walking control.
- the traditional humanoid robot walking control method is collectively referred to as a classical control method, which includes a zero moment point control method, a joint trajectory control method, and the like.
- This type of control method is calculated based on a large number of robot attitudes to obtain accurate position control and achieve stable walking of the robot. They are widely used in a variety of new humanoid robots such as ASIMO, HRP, HUBO and ATLAS.
- ASIMO, HRP, HUBO and ATLAS new humanoid robots
- HUBO HUBO
- ATLAS new humanoid robots
- Such control methods have inevitable shortcomings.
- the robot has to perform a large number of operations to achieve attitude stabilization, which results in a huge computational burden and reduces control efficiency.
- the trajectory control is difficult to achieve a humanoid walking mode, and most of the robots exhibit a rigid motion posture, which can be significantly improved.
- the parameter settings will directly affect the attitude control performance of the robot.
- the usual parameter setting process is based on manual comparison of a large number of robot experimental data and human body data. However, this process is long and complicated and cannot meet the requirements of automatic control.
- the parameters of manual debugging cannot meet the stable control of the robot under complex conditions, such as the difficult walking mode of the robot shifting and walking across obstacles.
- manual debugging parameters do not allow for automatic optimization of performance, stability, and adaptability.
- the robot under the traditional control method is far from the actual demand. It can be seen that the humanoid robots at this stage still have shortcomings such as low stability, low efficiency and low adaptability in attitude control.
- the present invention is directed to the above problems in the prior art, and proposes a dynamic walking control system for a bionic biped robot based on particle swarm optimization and reinforcement learning algorithm.
- the technical solutions adopted are as follows:
- the dynamic walking control system includes N sets of motion control units, wherein N is an integer greater than 1;
- the motion control unit includes a motion mode control center, a central mode control center, an action phase control center, an attitude reflection control center, and a joint unit
- the modulation signal output end of the motion mode control center is connected to the modulation signal input end of the joint unit;
- the stimulation signal output end of the motion mode control center is connected to the stimulation signal input end of the central mode control center;
- the center excitation signal output ends are respectively connected to the excitation signal input end of the action phase control center and the excitation signal input end of the joint unit;
- the action signal output end of the action phase control center is connected to the action signal input end of the attitude reflection control center.
- the dynamic walking control system further includes a balance sense control center and an electromechanical integrated unit; the signal output end of the balance sense control center is connected to the attitude signal input end of the attitude reflection control center; and the signal of the balance sense control center The input end is connected to the attitude signal output end of the electromechanical integrated unit; the somatosensory signal output end of the electromechanical integrated unit is connected to the somatosensory signal input end of the joint unit; the control signal input end of the electromechanical integrated unit and the control signal output of the joint unit Connected to the end.
- the electromechanical integration unit comprises an inertial measurement system, a somatosensory system and a joint control module; the signal output end of the inertial measurement system is an attitude signal output end of the electromechanical unit; the signal output end of the somatosensory system is The somatosensory signal output end of the electromechanical unit; the control signal input end of the joint control module is the control signal input end of the electromechanical unit.
- the joint unit includes a sensor unit; the signal output ends of the sensor unit are respectively connected to a motion mode control center, a central mode control center, an attitude reflection control center, and a sensing signal input end of the driving excitation controller;
- the signal input of the sensor unit is the somatosensory signal input of the joint unit.
- the joint unit further includes an underlying reflection controller and a driving excitation controller; an excitation signal input end of the bottom reflection controller and an excitation signal input end of the driving excitation controller are an excitation signal input end of the joint unit; The position signal output end of the bottom reflection controller and the torque control signal output end of the drive excitation controller are both connected to the control signal input end of the electromechanical unit.
- the bottom reflection controller includes an upper body stabilization controller and a locking hip joint controller.
- the drive excitation controller includes a swing leg controller, a front sill controller, a swing arm controller, and a knee bend controller.
- the action phase control center includes a support phase, a promotion phase, a stabilization phase, a swing leg phase, and a landing phase.
- the walking control system of the invention divides the robot walking process into five stages, can realize the stable control of the biped robot walking, simplifies the control process, and greatly improves the motion stability of the biped robot.
- the walking control system of the present invention adopts a 4BLC system design based on bionics control, has a multi-level stereo control structure, does not rely on traditional control algorithms such as zmp in the control process, does not require complex robot modeling, and omits cumbersome The computing burden.
- the travel control system of the present invention adopts dual control of position control and torque control in active control, which is a higher precision of motion control of the robot.
- the walking control system of the present invention employs passive control on multiple joints of the biped robot, the robot walks The process will be more human-like and more energy efficient.
- FIG. 1 is a schematic view showing the overall structure of a 4BLC system according to the present invention.
- FIG. 2 is a schematic structural diagram of a single group dynamic control unit according to the present invention.
- FIG. 3 is a schematic diagram of five behavioral stages included in the action phase control center of the present invention.
- FIG. 5 is a schematic diagram of the operation principle of the bottom layer emission controller and the driving excitation controller excited by the support phase in the present invention
- FIG. 6 is a schematic diagram showing the operation principle of the bottom emission controller and the driving excitation controller excited by the propulsion phase according to the present invention
- FIG. 7 is a schematic diagram showing the operation principle of the bottom layer emission controller and the driving excitation controller excited in the stable phase according to the present invention.
- FIG. 8 is a schematic diagram showing the operation principle of the bottom layer emission controller and the driving excitation controller excited by the swing leg stage according to the present invention
- FIG. 9 is a schematic diagram showing the operation principle of the underlying launch controller and the drive excitation controller excited by the landing phase of the present invention.
- the bionic biped robot dynamic walking control system has a multi-level control structure of the 4BLC system, that is, a bionics control that replaces the traditional biped robot control method is used in the multi-level structure 4BLC system.
- This biomimetic control approach applies numerous biological and biodynamic findings to the control philosophy of biped robots.
- FIG. 1 is a schematic structural view of a 4BLC system according to the present invention
- FIG. 2 is a schematic structural view of a single-group dynamic control unit according to the present invention
- the overall structure of the control system proposed by the present invention can be seen in conjunction with FIG. 1 and FIG.
- the dynamic walking control system includes N sets of motion control units, wherein N is an integer greater than 1; each set of motion control units includes 6 levels of control structures, respectively, the highest level mimicking the human brain's motion mode control center, imitating the human spinal nerve
- the modulation signal output end of the motion mode control center is connected to the modulation signal input end of the joint unit; the stimulation signal output end of the motion mode control center is connected to the stimulation signal input end of the central mode control center; and the excitation signal output of the central mode control center is The terminals are respectively connected to the excitation signal input end of the action phase control center and the excitation signal input end of the joint unit; the action signal output end of the action phase control center is connected to the action signal input end of the attitude reflection control center.
- the dynamic walking control system further comprises a balance sense control center and an electromechanical integrated unit; the signal output end of the balance sense control center is connected with the attitude signal input end of the attitude reflection control center; the signal input end of the balance sense control center is integrated with the electromechanical unit The attitude signal output end of the element is connected; the somatosensory signal output end of the electromechanical unit is connected to the somatosensory signal input end of the joint unit; the control signal input end of the electromechanical unit is connected to the control signal output end of the joint unit.
- the electromechanical integrated unit comprises an inertial measurement system, a somatosensory system and a joint control module; the signal output end of the inertial measurement system is the attitude signal output end of the electromechanical integrated unit; the signal output end of the somatosensory system is the somatosensory signal output end of the electromechanical integrated unit; The control signal input end of the joint control module is a control signal input end of the electromechanical integrated unit.
- the joint unit includes a sensor unit; the signal output ends of the sensor unit are respectively connected to the motion mode control center, the central mode control center, the attitude reflection control center, and the sensing signal input end of the driving excitation controller; the signal input end of the sensor unit is a joint The somatosensory signal input of the unit.
- the joint unit further includes an underlying reflection controller and a driving excitation controller; the excitation signal input end of the bottom reflection controller and the excitation signal input end of the driving excitation controller are the excitation signal input end of the joint unit; the position signal of the bottom reflection controller Both the output and the torque control signal output of the drive excitation controller are connected to the control signal input of the electromechanical unit.
- the bottom reflection controller includes an upper body stability controller and a lock hip joint controller;
- the drive excitation controller includes a swing leg controller, a front sill controller, a swing arm controller, and a knee bend controller.
- the action phase control center includes a support phase, a promotion phase, a stabilization phase, a swing leg phase, and a landing phase.
- the motion mode control center mimics the operating mechanism of the human brain, giving the entire control system commands for different motion modes, such as standing mode, walking mode, and running mode. When a certain motion mode is selected, the corresponding stimulus signal is passed to the central mode control center in the next layer.
- the central mode control center mimics the mechanism of the human central nervous system. Its role is to establish a link between the corresponding behavioral phase control center and the motion mode control center. When receiving the stimulus signal from the upper motion mode control center, the central mode control center will generate the corresponding excitation signal, allowing the system to select the behavior phase of the robot.
- the action phase control center is used to select the behavior phase of the robot walking process. During the walking process, the robot gradually experiences each phase in sequence and continuously forms a cycle. The transformation of different phases is determined by the attitude and excitation event of the robot.
- the stage process is shown in Figure 3.
- the attitude reflection controller that mimics the human reflection arc is a feedback controller that intervenes and adjusts the attitude of the robot inside the system. Different attitude reflection controllers will be used for different purposes, the main responsibility is to ensure the direction of the upper body, the position control of the robot and the speed control.
- the bottom reflection controller is the position control and torque control of the robot joint. Different underlying reflection controllers will be fired at different stages of behavior.
- the drive excitation controller is a feedforward controller that mimics the human muscle setting and will produce a shape similar to the Gaussian function.
- the amount of torque produced by the torque can be expressed as follows:
- T 1 represents the maximum torque start time parameter
- T 2 is the maximum torque end time parameter
- T 3 is the torque action time parameter
- the torque curve is shown in FIG. 4 .
- Action Phase Control Center can control the robot to select five different behavioral phases, including the support phase, the propulsion phase, the stabilization phase, the swing leg phase, and the landing phase.
- the specific process of walking is:
- FIG. 5 to 9 show the controllers in the five stages of support, propulsion, stability, swinging legs and landing. The situation that was triggered:
- the bottom-level reflection controller upper body stability controller (Stabilize Pelvis) is excited.
- the controller controls the activity on the hip joint of the support leg, and the support leg hip joint is locked in the lateral direction to ensure the upper body is stable.
- the attitude reflection control center includes: the forward speed controller, the side up controller, and the lateral stability controller are activated.
- the upper body stability controller continues to be activated.
- the swinging leg's drive excitation controller leg swinging and the support leg's drive excitation controller leg propel are activated.
- the swing leg controller produces torque on the hip joint of the swinging leg, creating an action of swinging the forward leg.
- the front stern controller produces torque on the ankle joint, producing a forward driving force.
- the attitude reflection control center includes: the forward speed controller, the side up controller, and the lateral stability controller are activated.
- the attitude reflection control center includes: the lateral stability controller is activated.
- the swing arm controller and the swing leg controller will continue to fire.
- the underlying reflection controller locks the hip hip to be activated, to prevent the leg from swinging, and to maintain the hip joint in an appropriate The angle is good for the feet to land.
- the locking hip joint controller will continue to be excited.
- the drive stimulus controller weight acceptance will be activated.
- the knee flexion controller will produce torque control and position control on the knee joint, and the knee joint will be properly bent to reduce the impact of the ground on the body.
Landscapes
- Engineering & Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Combustion & Propulsion (AREA)
- Transportation (AREA)
- Mechanical Engineering (AREA)
- Manipulator (AREA)
Abstract
Description
本发明涉及基于粒子群优化和强化学习算法的仿生双足机器人动态行走控制系统，属于仿生人形机器人行走控制技术领域。The invention relates to a bionic biped robot dynamic walking control system based on particle swarm optimization and reinforcement learning algorithm, and belongs to the technical field of bionic humanoid robot walking control.
传统的人形机器人行走控制方法被统称为经典控制方法，这其中包括零力矩点(zero moment point)控制方法、关节轨迹控制方法等。这类控制方法根据大量的机器人姿态计算，以得到精确的位置控制，实现机器人的稳定行走。它们被广泛的应用在各类新型的人形走形机器人上面，例如ASIMO，HRP，HUBO和ATLAS等机器人。然而，这类控制方法存在不可避免的缺点。首先，机器人必须要进行大量的运算以实现姿态稳定，由此产生巨大的运算负担，降低了控制效率。其次，轨迹控制难以实现类人型的行走模式，机器人大多表现出僵硬的运动姿态，能熬显著提高。再者，由于这类控制系统内部的前馈和反馈控制器中包括大量的参数，而参数的设定将直接影响机器人的姿态控制性能。往常的参数设定过程是根据大量的机器人实验数据和人体数据的对比后，经行的手工调试。然而，这个过程漫长而复杂，不能满足自动控制的要求。其次，手工调试的参数无法满足复杂条件下机器人的稳定控制，例如机器人变速行走、跨过障碍物行走等高难度行走模式。再者，手工调试参数无法实现对于效能、稳定性和适应性的自动优化。The traditional humanoid robot walking control method is collectively referred to as a classical control method, which includes a zero moment point control method, a joint trajectory control method, and the like. This type of control method is calculated based on a large number of robot attitudes to obtain accurate position control and achieve stable walking of the robot. They are widely used in a variety of new humanoid robots such as ASIMO, HRP, HUBO and ATLAS. However, such control methods have inevitable shortcomings. First of all, the robot has to perform a large number of operations to achieve attitude stabilization, which results in a huge computational burden and reduces control efficiency. Secondly, the trajectory control is difficult to achieve a humanoid walking mode, and most of the robots exhibit a rigid motion posture, which can be significantly improved. Furthermore, since the feedforward and feedback controllers inside such control systems include a large number of parameters, the parameter settings will directly affect the attitude control performance of the robot. The usual parameter setting process is based on manual comparison of a large number of robot experimental data and human body data. However, this process is long and complicated and cannot meet the requirements of automatic control. Secondly, the parameters of manual debugging cannot meet the stable control of the robot under complex conditions, such as the difficult walking mode of the robot shifting and walking across obstacles. Furthermore, manual debugging parameters do not allow for automatic optimization of performance, stability, and adaptability.
由此，传统的控制方法下的机器人与实际需求相距甚远。由此可见，现阶段的人形机器人在姿态控制上仍然存在稳定性低、效率低、适应性能低等缺点。Therefore, the robot under the traditional control method is far from the actual demand. It can be seen that the humanoid robots at this stage still have shortcomings such as low stability, low efficiency and low adaptability in attitude control.
发明内容Summary of the invention
本发明针对上述现有技术中存在的问题，提出了一种基于粒子群优化和强化学习算法的仿生双足机器人动态行走控制系统，所采取的技术方案如下：The present invention is directed to the above problems in the prior art, and proposes a dynamic walking control system for a bionic biped robot based on particle swarm optimization and reinforcement learning algorithm. The technical solutions adopted are as follows:
所述动态行走控制系统包括N组动作控制单元，其中，N为大于1的整数；所述动作控制单元包括运动模式控制中心、中枢模式控制中心、行动阶段控制中心、姿态反射控制中心和联合单元；所述运动模式控制中心的调制信号输出端与联合单元的调制信号输入端相连；所述运动模式控制中心的刺激信号输出端与中枢模式控制中心的刺激信号输入端相连；所述中枢模式控制中心的激励信号输出端分别与行动阶段控制中心的激励信号输入端和联合单元的激励信号输入端相连；所述行动阶段控制中心的行动信号输出端与姿态反射控制中心的行动信号输入端相连。 The dynamic walking control system includes N sets of motion control units, wherein N is an integer greater than 1; the motion control unit includes a motion mode control center, a central mode control center, an action phase control center, an attitude reflection control center, and a joint unit The modulation signal output end of the motion mode control center is connected to the modulation signal input end of the joint unit; the stimulation signal output end of the motion mode control center is connected to the stimulation signal input end of the central mode control center; the central mode control The center excitation signal output ends are respectively connected to the excitation signal input end of the action phase control center and the excitation signal input end of the joint unit; the action signal output end of the action phase control center is connected to the action signal input end of the attitude reflection control center.
优先地，所述动态行走控制系统还包括平衡感控制中心、机电一体单元；所述平衡感控制中心的信号输出端与姿态反射控制中心的姿态信号输入端相连；所述平衡感控制中心的信号输入端与机电一体单元的姿态信号输出端相连；所述机电一体单元的体感信号输出端与联合单元的体感信号输入端相连；所述机电一体单元的控制信号输入端与联合单元的控制信号输出端相连。Preferably, the dynamic walking control system further includes a balance sense control center and an electromechanical integrated unit; the signal output end of the balance sense control center is connected to the attitude signal input end of the attitude reflection control center; and the signal of the balance sense control center The input end is connected to the attitude signal output end of the electromechanical integrated unit; the somatosensory signal output end of the electromechanical integrated unit is connected to the somatosensory signal input end of the joint unit; the control signal input end of the electromechanical integrated unit and the control signal output of the joint unit Connected to the end.
优先地，所述机电一体单元包括惯性测量系统、体感系统和联合控制模块；所述惯性测量系统的信号输出端即为机电一体单元的姿态信号输出端；所述体感系统的信号输出端即为机电一体单元的体感信号输出端；所述联合控制模块的控制信号输入端即为机电一体单元的控制信号输入端。Preferably, the electromechanical integration unit comprises an inertial measurement system, a somatosensory system and a joint control module; the signal output end of the inertial measurement system is an attitude signal output end of the electromechanical unit; the signal output end of the somatosensory system is The somatosensory signal output end of the electromechanical unit; the control signal input end of the joint control module is the control signal input end of the electromechanical unit.
优先地，所述联合单元包括传感器单元；所述传感器单元的信号输出端分别与运动模式控制中心、中枢模式控制中心、姿态反射控制中心和驱动激励控制器的传感信号输入端相连；所述传感器单元的信号输入端即为联合单元的体感信号输入端。Preferentially, the joint unit includes a sensor unit; the signal output ends of the sensor unit are respectively connected to a motion mode control center, a central mode control center, an attitude reflection control center, and a sensing signal input end of the driving excitation controller; The signal input of the sensor unit is the somatosensory signal input of the joint unit.
优先地，所述联合单元还包括底层反射控制器和驱动激励控制器；所述底层反射控制器的激励信号输入端和驱动激励控制器的激励信号输入端即为联合单元的激励信号输入端；所述底层反射控制器的位置信号输出端和驱动激励控制器的扭矩控制信号输出端均与机电一体单元的控制信号输入端相连。Preferably, the joint unit further includes an underlying reflection controller and a driving excitation controller; an excitation signal input end of the bottom reflection controller and an excitation signal input end of the driving excitation controller are an excitation signal input end of the joint unit; The position signal output end of the bottom reflection controller and the torque control signal output end of the drive excitation controller are both connected to the control signal input end of the electromechanical unit.
优选地，所述底层反射控制器包括上身稳定控制器和锁止髋关节控制器。Preferably, the bottom reflection controller includes an upper body stabilization controller and a locking hip joint controller.
优选地，所述驱动激励控制器包括摆腿控制器、前蹬控制器、摆臂控制器和屈膝控制器。Preferably, the drive excitation controller includes a swing leg controller, a front sill controller, a swing arm controller, and a knee bend controller.
优先地，所述行动阶段控制中心包括支撑阶段、推进阶段、稳定阶段、摆腿阶段和着地阶段。Preferentially, the action phase control center includes a support phase, a promotion phase, a stabilization phase, a swing leg phase, and a landing phase.
本发明有益效果：The beneficial effects of the invention:
1.本发明所述行走控制系统将机器人行走过程分为5个阶段，可以实现双足机器人行走的稳定控制，简化了控制过程，大幅度的提高了双足机器人的运动稳定性。1. The walking control system of the invention divides the robot walking process into five stages, can realize the stable control of the biped robot walking, simplifies the control process, and greatly improves the motion stability of the biped robot.
2.本发明所述行走控制系统采用基于仿生学控制的4BLC系统设计，具有多层次立体式控制结构，在控制过程中不依赖zmp等传统控制算法，不需要复杂机器人建模，省略了繁琐的运算负担。2. The walking control system of the present invention adopts a 4BLC system design based on bionics control, has a multi-level stereo control structure, does not rely on traditional control algorithms such as zmp in the control process, does not require complex robot modeling, and omits cumbersome The computing burden.
3.本发明所述行走控制系统在主动控制方面采用了位置控制和扭矩控制的双重控制，是机器人的运动控制精度更高。3. The travel control system of the present invention adopts dual control of position control and torque control in active control, which is a higher precision of motion control of the robot.
4.由于本发明所述行走控制系统在双足机器人的多个关节上采用了被动控制，机器人行走 过程将更加类人化，能耗效率更高。4. Since the walking control system of the present invention employs passive control on multiple joints of the biped robot, the robot walks The process will be more human-like and more energy efficient.
图1为本发明所述4BLC系统的总体结构示意图；1 is a schematic view showing the overall structure of a 4BLC system according to the present invention;
图2为本发明所述单组动态控制单元的结构示意图；2 is a schematic structural diagram of a single group dynamic control unit according to the present invention;
图3为本发明所述行动阶段控制中心包括的5个行为阶段示意图；3 is a schematic diagram of five behavioral stages included in the action phase control center of the present invention;
图4为本发明所述驱动激励控制器输出各形态的扭矩曲线图；4 is a torque curve diagram of various forms of driving excitation controller output according to the present invention;
图5为本发明所述支撑阶段激发的底层发射控制器和驱动激励控制器运行原理示意图；FIG. 5 is a schematic diagram of the operation principle of the bottom layer emission controller and the driving excitation controller excited by the support phase in the present invention; FIG.
图6为本发明所述推进阶段激发的底层发射控制器和驱动激励控制器运行原理示意图；6 is a schematic diagram showing the operation principle of the bottom emission controller and the driving excitation controller excited by the propulsion phase according to the present invention;
图7为本发明所述稳定阶段激发的底层发射控制器和驱动激励控制器运行原理示意图；7 is a schematic diagram showing the operation principle of the bottom layer emission controller and the driving excitation controller excited in the stable phase according to the present invention;
图8为本发明所述摆腿阶段激发的底层发射控制器和驱动激励控制器运行原理示意图；8 is a schematic diagram showing the operation principle of the bottom layer emission controller and the driving excitation controller excited by the swing leg stage according to the present invention;
图9为本发明所述着地阶段激发的底层发射控制器和驱动激励控制器运行原理示意图。FIG. 9 is a schematic diagram showing the operation principle of the underlying launch controller and the drive excitation controller excited by the landing phase of the present invention.
下面结合具体实施例对本发明做进一步说明，但本发明不受实施例的限制。The invention is further illustrated by the following specific examples, but the invention is not limited by the examples.
该仿生双足机器人动态行走控制系统具有多层次控制结构的4BLC系统，即在多层次结构的4BLC系统中运用了一种代替传统双足机器人控制方法的仿生学控制。这种仿生学控制方法将众多生物学和生物动力学的发现应用到双足机器人的控制理念。图1为本发明所述4BLC系统的总体结构示意图；图2为本发明所述单组动态控制单元的结构示意图；结合图1和图2可以看出本发明提出的控制系统的整体结构，该动态行走控制系统包括N组动作控制单元，其中，N为大于1的整数；每组动作控制单元包括6个层次的控制结构，分别是最高层模仿人类大脑的运动模式控制中心、模仿人类脊柱神经的中枢模式控制中心、行动阶段控制中心、模仿人类反射弧的姿态反射控制器和底层反射控制器、模仿人类肌肉的驱动激励控制器；其中，底层反射控制器和驱动激励控制器整合为一个联合单元。其中，The bionic biped robot dynamic walking control system has a multi-level control structure of the 4BLC system, that is, a bionics control that replaces the traditional biped robot control method is used in the multi-level structure 4BLC system. This biomimetic control approach applies numerous biological and biodynamic findings to the control philosophy of biped robots. 1 is a schematic structural view of a 4BLC system according to the present invention; FIG. 2 is a schematic structural view of a single-group dynamic control unit according to the present invention; and the overall structure of the control system proposed by the present invention can be seen in conjunction with FIG. 1 and FIG. The dynamic walking control system includes N sets of motion control units, wherein N is an integer greater than 1; each set of motion control units includes 6 levels of control structures, respectively, the highest level mimicking the human brain's motion mode control center, imitating the human spinal nerve The central mode control center, the action phase control center, the attitude reflection controller and the bottom reflection controller that mimic the human reflex arc, and the drive excitation controller that mimics the human muscle; wherein the bottom reflection controller and the drive excitation controller are integrated into one joint unit. among them,
同时，运动模式控制中心的调制信号输出端与联合单元的调制信号输入端相连；运动模式控制中心的刺激信号输出端与中枢模式控制中心的刺激信号输入端相连；中枢模式控制中心的激励信号输出端分别与行动阶段控制中心的激励信号输入端和联合单元的激励信号输入端相连；行动阶段控制中心的行动信号输出端与姿态反射控制中心的行动信号输入端相连。Meanwhile, the modulation signal output end of the motion mode control center is connected to the modulation signal input end of the joint unit; the stimulation signal output end of the motion mode control center is connected to the stimulation signal input end of the central mode control center; and the excitation signal output of the central mode control center is The terminals are respectively connected to the excitation signal input end of the action phase control center and the excitation signal input end of the joint unit; the action signal output end of the action phase control center is connected to the action signal input end of the attitude reflection control center.
动态行走控制系统还包括平衡感控制中心、机电一体单元；平衡感控制中心的信号输出端与姿态反射控制中心的姿态信号输入端相连；平衡感控制中心的信号输入端与机电一体单 元的姿态信号输出端相连；机电一体单元的体感信号输出端与联合单元的体感信号输入端相连；机电一体单元的控制信号输入端与联合单元的控制信号输出端相连。机电一体单元包括惯性测量系统、体感系统和联合控制模块；惯性测量系统的信号输出端即为机电一体单元的姿态信号输出端；体感系统的信号输出端即为机电一体单元的体感信号输出端；所述联合控制模块的控制信号输入端即为机电一体单元的控制信号输入端。The dynamic walking control system further comprises a balance sense control center and an electromechanical integrated unit; the signal output end of the balance sense control center is connected with the attitude signal input end of the attitude reflection control center; the signal input end of the balance sense control center is integrated with the electromechanical unit The attitude signal output end of the element is connected; the somatosensory signal output end of the electromechanical unit is connected to the somatosensory signal input end of the joint unit; the control signal input end of the electromechanical unit is connected to the control signal output end of the joint unit. The electromechanical integrated unit comprises an inertial measurement system, a somatosensory system and a joint control module; the signal output end of the inertial measurement system is the attitude signal output end of the electromechanical integrated unit; the signal output end of the somatosensory system is the somatosensory signal output end of the electromechanical integrated unit; The control signal input end of the joint control module is a control signal input end of the electromechanical integrated unit.
联合单元包括传感器单元；传感器单元的信号输出端分别与运动模式控制中心、中枢模式控制中心、姿态反射控制中心和驱动激励控制器的传感信号输入端相连；传感器单元的信号输入端即为联合单元的体感信号输入端。联合单元还包括底层反射控制器和驱动激励控制器；底层反射控制器的激励信号输入端和驱动激励控制器的激励信号输入端即为联合单元的激励信号输入端；底层反射控制器的位置信号输出端和驱动激励控制器的扭矩控制信号输出端均与机电一体单元的控制信号输入端相连。同时，底层反射控制器包括上身稳定控制器和锁止髋关节控制器；驱动激励控制器包括摆腿控制器、前蹬控制器、摆臂控制器和屈膝控制器。行动阶段控制中心包括支撑阶段、推进阶段、稳定阶段、摆腿阶段和着地阶段。The joint unit includes a sensor unit; the signal output ends of the sensor unit are respectively connected to the motion mode control center, the central mode control center, the attitude reflection control center, and the sensing signal input end of the driving excitation controller; the signal input end of the sensor unit is a joint The somatosensory signal input of the unit. The joint unit further includes an underlying reflection controller and a driving excitation controller; the excitation signal input end of the bottom reflection controller and the excitation signal input end of the driving excitation controller are the excitation signal input end of the joint unit; the position signal of the bottom reflection controller Both the output and the torque control signal output of the drive excitation controller are connected to the control signal input of the electromechanical unit. Meanwhile, the bottom reflection controller includes an upper body stability controller and a lock hip joint controller; the drive excitation controller includes a swing leg controller, a front sill controller, a swing arm controller, and a knee bend controller. The action phase control center includes a support phase, a promotion phase, a stabilization phase, a swing leg phase, and a landing phase.
在4BLC结构中，各控制中心及部件的作用如下：In the 4BLC structure, the functions of each control center and components are as follows:
1、运动模式控制中心模仿人类大脑的运行机理，给予整个控制系统发出不同的运动模式的命令，例如，站立模式、行走模式和奔跑模式。当某一个运动模式被选定后，相应的刺激信号将传递给下面一层的中枢模式控制中心里。1. The motion mode control center mimics the operating mechanism of the human brain, giving the entire control system commands for different motion modes, such as standing mode, walking mode, and running mode. When a certain motion mode is selected, the corresponding stimulus signal is passed to the central mode control center in the next layer.
2、中枢模式控制中心模仿人类中枢神经的机理，它的作用是建立相应的行为阶段控制中心和运动模式控制中心之间的联系。当接受来自上层运动模式控制中心的刺激信号时，中枢模式控制中心将产生于之对应的激励信号，让系统选择机器人所处的行为阶段。2. The central mode control center mimics the mechanism of the human central nervous system. Its role is to establish a link between the corresponding behavioral phase control center and the motion mode control center. When receiving the stimulus signal from the upper motion mode control center, the central mode control center will generate the corresponding excitation signal, allowing the system to select the behavior phase of the robot.
3、行动阶段控制中心用于选择机器人行走过程中的行为阶段，机器人在行走的过程中，是逐步按顺序经历各个阶段，并不断形成循环，不同阶段的转化是由机器人的姿态和激发事件决定的，其阶段过程如图3所示。3. The action phase control center is used to select the behavior phase of the robot walking process. During the walking process, the robot gradually experiences each phase in sequence and continuously forms a cycle. The transformation of different phases is determined by the attitude and excitation event of the robot. The stage process is shown in Figure 3.
4、模仿人类反射弧的姿态反射控制器是位于系统内部对机器人姿态进行干预和调整反馈控制器。不同的姿态反射控制器将用于不同的目的，主要职责是保证上半身的方向、机器人的位置控制和速度控制。4. The attitude reflection controller that mimics the human reflection arc is a feedback controller that intervenes and adjusts the attitude of the robot inside the system. Different attitude reflection controllers will be used for different purposes, the main responsibility is to ensure the direction of the upper body, the position control of the robot and the speed control.
5、底层反射控制器是对机器人关节进行位置控制和扭矩控制。在不同的行为阶段，不同的底层反射控制器将被激发。5. The bottom reflection controller is the position control and torque control of the robot joint. Different underlying reflection controllers will be fired at different stages of behavior.
6、驱动激励控制器是模仿人类肌肉设定的前馈式控制器，将产生类似于高斯函数外形的6. The drive excitation controller is a feedforward controller that mimics the human muscle setting and will produce a shape similar to the Gaussian function.
扭矩所产生的扭矩大小可以根据下式表达：The amount of torque produced by the torque can be expressed as follows:
其中A是最大扭矩参数，T1代表最大扭矩开始时间参数，T2是最大扭矩结束时间参数，T3是扭矩作用时间参数，该扭矩的曲线如图4所示。Where A is the maximum torque parameter, T 1 represents the maximum torque start time parameter, T 2 is the maximum torque end time parameter, and T 3 is the torque action time parameter, and the torque curve is shown in FIG. 4 .
此外，行动阶段控制中心可以控制机器人选择5种不同的行为阶段，包括支撑阶段、推进阶段、稳定阶段、摆腿阶段和着地阶段。其行走的具体过程为：In addition, the Action Phase Control Center can control the robot to select five different behavioral phases, including the support phase, the propulsion phase, the stabilization phase, the swing leg phase, and the landing phase. The specific process of walking is:
(1)、当摆动腿的膝关节完成了关节锁定(locked knee)，即膝关节处于伸直的状态时，支撑腿将从支撑阶段变为推进阶段，摆动腿将由摆动阶段变为着地阶段。(1) When the knee joint of the swinging leg completes the locked knee, that is, when the knee joint is in a straight state, the support leg will change from the support phase to the advancement phase, and the swing leg will change from the swing phase to the ground phase.
(2)、当摆动腿完成了脚跟触底的过程，脚步实现了脚掌全部接触(full contact)，机器人进入单腿支撑状态。此时，摆动腿将从着地阶段变为支撑阶段，而另一条支撑腿则继续进入稳定阶段。(2) When the swinging leg completes the process of bottoming the heel, the footstep realizes full contact of the sole of the foot, and the robot enters the state of single leg support. At this point, the swing leg will change from the landing phase to the support phase, while the other support leg will continue to enter the stabilization phase.
(3)、处于稳定阶段的支撑腿不断蹬地、伸展脚踝，给机器人提供向前的动力，直到脚趾离地(toe-off)完成。此时，处于稳定阶段的支撑腿过渡到摆动阶段，转变为摆动腿。另一条处于支撑腿则仍然处于支撑阶段。至此，机器人完成了两条腿5个阶段的不断转换，保证了稳定的行走。(3) The support legs in a stable phase continually squat and stretch the ankles to provide forward power to the robot until the toe is completed. At this time, the support legs in the stable phase transition to the swing phase and become the swing legs. The other leg is still in the support phase. At this point, the robot completed the continuous conversion of the two legs in five stages, ensuring stable walking.
在机器人行走过程中，机器人处在不同阶段，其相应的控制器会被逐次激发，图5至图9分别展示了机器人处于支撑、推进、稳定、摆腿和着地这5个阶段时各控制器被激发的情况：During the robot walking process, the robot is at different stages, and the corresponding controllers are excited successively. Figures 5 to 9 show the controllers in the five stages of support, propulsion, stability, swinging legs and landing. The situation that was triggered:
(1)、在支撑阶段，底层反射控制器上身稳定控制器(Stabilize Pelvis)被激发。此控制器控制支撑腿的髋关节上的活动，将支撑腿髋关节在横向方向上被锁死，保证上身稳定。姿态反射控制中心包括：前向速度控制器、侧向上身控制器和侧向稳定控制器被激发。(1) In the support phase, the bottom-level reflection controller upper body stability controller (Stabilize Pelvis) is excited. The controller controls the activity on the hip joint of the support leg, and the support leg hip joint is locked in the lateral direction to ensure the upper body is stable. The attitude reflection control center includes: the forward speed controller, the side up controller, and the lateral stability controller are activated.
(2)、在推进阶段，上身稳定控制器继续被激发。同时，摆动腿的驱动激励控制器摆腿控制器(leg swinging)和支撑腿的驱动激励控制器前蹬控制器(leg propel)被激发。摆腿控制器产生扭矩在摆动腿的髋关节上，产生向前摆腿的动作。前蹬控制器产生扭矩在脚踝关节上，产生向前驱动力。姿态反射控制中心包括：前向速度控制器、侧向上身控制器和侧向稳定控制器被激发。(2) During the advancement phase, the upper body stability controller continues to be activated. At the same time, the swinging leg's drive excitation controller leg swinging and the support leg's drive excitation controller leg propel are activated. The swing leg controller produces torque on the hip joint of the swinging leg, creating an action of swinging the forward leg. The front stern controller produces torque on the ankle joint, producing a forward driving force. The attitude reflection control center includes: the forward speed controller, the side up controller, and the lateral stability controller are activated.
(3)、在稳定阶段，摆腿控制器和前蹬控制器继续被激发。同时，驱动激励控制器摆臂控制器(arm swing)被激发，产生扭矩在肩关节，产生有规律的手臂摆动。姿态反射控制中心包括：侧向稳定控制器被激发。(3) In the steady phase, the swing leg controller and the front cymbal controller continue to be activated. At the same time, the drive excitation controller arm swing is activated, generating torque on the shoulder joint, producing a regular arm swing. The attitude reflection control center includes: the lateral stability controller is activated.
(4)、在摆腿阶段，摆臂控制器和摆腿控制器将持续激发。另外，在摆腿的末期，底层反射控制器锁止髋关节(lock hip)将被激发，用于阻止腿部摆动，并维持髋关节在一个适当 的角度，利于脚部着地。(4) In the swinging stage, the swing arm controller and the swing leg controller will continue to fire. In addition, at the end of the swing leg, the underlying reflection controller locks the hip hip to be activated, to prevent the leg from swinging, and to maintain the hip joint in an appropriate The angle is good for the feet to land.
(5)在着地阶段，锁止髋关节控制器将延续激。另外，驱动激励控制器屈膝控制器(weight acceptance)将被激发。在着地的过程中，屈膝控制器将对膝关节产生扭矩控制和位置控制，膝关节适当弯曲，减少着地对身体的冲击。(5) At the landing stage, the locking hip joint controller will continue to be excited. In addition, the drive stimulus controller weight acceptance will be activated. In the process of landing, the knee flexion controller will produce torque control and position control on the knee joint, and the knee joint will be properly bent to reduce the impact of the ground on the body.
虽然本发明已以较佳的实施例公开如上，但其并非用以限定本发明，任何熟悉此技术的人，在不脱离本发明的精神和范围内，都可以做各种改动和修饰，因此本发明的保护范围应该以权利要求书所界定的为准。 Although the present invention has been disclosed in the above preferred embodiments, it is not intended to limit the invention, and various modifications and changes can be made thereto without departing from the spirit and scope of the invention. The scope of the invention should be determined by the scope of the claims.
Claims (8)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| PCT/CN2016/079543 WO2017181319A1 (en) | 2016-04-18 | 2016-04-18 | Particle swarm optimization and reinforcement learning algorithm-based dynamic walking control system for biomimetic biped robot |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| PCT/CN2016/079543 WO2017181319A1 (en) | 2016-04-18 | 2016-04-18 | Particle swarm optimization and reinforcement learning algorithm-based dynamic walking control system for biomimetic biped robot |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2017181319A1 true WO2017181319A1 (en) | 2017-10-26 |
Family
ID=60115553
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2016/079543 Ceased WO2017181319A1 (en) | 2016-04-18 | 2016-04-18 | Particle swarm optimization and reinforcement learning algorithm-based dynamic walking control system for biomimetic biped robot |
Country Status (1)
| Country | Link |
|---|---|
| WO (1) | WO2017181319A1 (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111290275A (en) * | 2020-02-23 | 2020-06-16 | 浙江科技学院 | Optimal control method of sewage treatment based on reinforcement learning particle swarm algorithm |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1290590A (en) * | 1999-09-20 | 2001-04-11 | 索尼公司 | Robot walking controller and walking control method |
| CN201208992Y (en) * | 2008-04-24 | 2009-03-18 | 上海电气集团股份有限公司 | Control system for two feet robot walking mechanism |
| CN101751037A (en) * | 2008-12-03 | 2010-06-23 | 上海电气集团股份有限公司 | Dynamic walking control method for biped walking robot |
| CN102202613A (en) * | 2008-09-04 | 2011-09-28 | Iwalk股份有限公司 | Hybrid terrain-adaptive lower-extremity systems |
| EP2574527A2 (en) * | 2011-09-28 | 2013-04-03 | Samsung Electronics Co., Ltd | Robot and control method thereof |
| CN105446345A (en) * | 2015-07-02 | 2016-03-30 | 浙江大学 | Control system of humanoid biped robot |
-
2016
- 2016-04-18 WO PCT/CN2016/079543 patent/WO2017181319A1/en not_active Ceased
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1290590A (en) * | 1999-09-20 | 2001-04-11 | 索尼公司 | Robot walking controller and walking control method |
| CN201208992Y (en) * | 2008-04-24 | 2009-03-18 | 上海电气集团股份有限公司 | Control system for two feet robot walking mechanism |
| CN102202613A (en) * | 2008-09-04 | 2011-09-28 | Iwalk股份有限公司 | Hybrid terrain-adaptive lower-extremity systems |
| CN101751037A (en) * | 2008-12-03 | 2010-06-23 | 上海电气集团股份有限公司 | Dynamic walking control method for biped walking robot |
| EP2574527A2 (en) * | 2011-09-28 | 2013-04-03 | Samsung Electronics Co., Ltd | Robot and control method thereof |
| CN105446345A (en) * | 2015-07-02 | 2016-03-30 | 浙江大学 | Control system of humanoid biped robot |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111290275A (en) * | 2020-02-23 | 2020-06-16 | 浙江科技学院 | Optimal control method of sewage treatment based on reinforcement learning particle swarm algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8204626B2 (en) | Control device for mobile body | |
| CN100569579C (en) | A powered walking method for a biped robot | |
| US8417382B2 (en) | Control device for legged mobile body | |
| US8311677B2 (en) | Control device for legged mobile robot | |
| CN101531006B (en) | Power type walking method for biped robot | |
| Anderson et al. | Powered bipeds based on passive dynamic principles | |
| Deng et al. | Level-ground walking for a bipedal robot with a torso via hip series elastic actuators and its gait bifurcation control | |
| CN109760761A (en) | A kind of quadruped robot motion control method based on bionic principle and intuition | |
| CN106828654A (en) | A kind of four-leg bionic robot | |
| CN110405763B (en) | Planning method for multi-joint cooperative outbreak jump of humanoid robot | |
| CN101751037A (en) | Dynamic walking control method for biped walking robot | |
| Zhang et al. | Effects of head and tail as swinging appendages on the dynamic walking performance of a quadruped robot | |
| Wang et al. | Passive dynamic walking with flat feet and ankle compliance | |
| KR20200099664A (en) | Walking control system and control method of robot | |
| CN107315346B (en) | Humanoid robot gait planning method based on CPG model | |
| CN101898594B (en) | Walking method for dynamic biped robot | |
| CN104656440B (en) | Humanoid robot gait's generation method | |
| WO2017181319A1 (en) | Particle swarm optimization and reinforcement learning algorithm-based dynamic walking control system for biomimetic biped robot | |
| Wang et al. | Human-like ZMP generator and walking stabilizer based on divergent component of motion | |
| Huang et al. | Modeling and gait selection of passivity-based seven-link bipeds with dynamic series of walking phases | |
| Song et al. | CPG-based control design for bipedal walking on unknown slope surfaces | |
| Yan et al. | Human-in-the-loop optimization control for the ankle exoskeleton during walking based on iterative learning and particle swarm optimization algorithm | |
| Agarwal et al. | Study of toe joints to enhance locomotion of humanoid robots | |
| Ogino et al. | Learning energy-efficient walking with ballistic walking | |
| Fu et al. | A walking control strategy combining global sensory reflex and leg synchronization |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 16898909 Country of ref document: EP Kind code of ref document: A1 |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 16898909 Country of ref document: EP Kind code of ref document: A1 |