Man-machine cooperative control method for heavy-load rescue robot
By introducing a secure reinforcement learning model and human-machine collaborative control method in the heavy-load rescue robot, the problem of traditional methods ignoring instantaneous safety is solved, and more efficient and safe operation operations are achieved.
Patent Information
- Application Number
- CN202510090423.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-21
- Publication Date
- 2025-04-22
AI Technical Summary
Traditional reinforcement learning methods ignore instantaneous safety during the execution of heavy-load rescue robot tasks, resulting in potential dangers, affecting operational reliability and task success rate.
A human-machine collaborative control method for heavy-load rescue robots is proposed. By obtaining real-time data, a reward function of the security reinforcement learning model is designed, and a security reinforcement learning model based on deep deterministic strategy gradient algorithm is constructed. The control weight between the operator and the intelligent system is dynamically adjusted to realize human-machine collaborative control between the three.
Effectively ensure the operational safety of heavy-load rescue robots, improve operational efficiency in emergency rescue environments, and ensure operational reliability and task success rate.
Smart Images
Figure CN119861720A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of robot control technology, and in particular to a human-machine collaborative control method for a heavy-load rescue robot. Background Art
[0002] In emergency rescue environments, heavy-duty rescue robots face significant operational risks during mission execution due to the undulating terrain, steep slopes, and complex obstacle distribution. To meet these challenges and improve the robot's adaptability and mission completion efficiency in complex environments, researchers have begun to try to develop intelligent algorithms that adapt to dynamic environments as a solution.
[0003] Among them, reinforcement learning, as an intelligent method that can accumulate experience and optimize decision-making through interaction with the environment, has been widely used in algorithm design in this field. However, traditional reinforcement learning methods usually aim to maximize cumulative rewards. Although they can significantly improve long-term strategy benefits, they often ignore the instantaneous safety during task execution. This defect may lead to potential dangers in actual rescue missions, affecting the reliability of robot operations and the success rate of tasks. Traditional reinforcement learning methods rely on repeated trial and error interactions between heavy-duty rescue robots and the environment to accumulate experience and adapt to complex environments through dynamic decision-making. However, they tend to choose strategies with the largest cumulative rewards and ignore instantaneous safety issues, which may lead to dangerous behaviors in actual rescue. For example, in collapsed ruins, when a heavy-duty rescue robot grabs heavy objects, it may ignore the slope of the terrain or the distance to obstacles because the reinforcement learning strategy tends to maximize rewards, resulting in collisions between the mobile chassis and surrounding obstacles. Summary of the invention
[0004] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.
[0005] Therefore, the purpose of the present invention is to provide a human-machine collaborative control method for a heavy-loaded rescue robot to achieve three-way combined human-machine collaborative control, which can not only ensure the operation safety of the heavy-loaded rescue robot to a limited extent through a joint safety model, but also improve the operation efficiency of the heavy-loaded rescue robot in an emergency rescue environment through the efficient collaboration of the three.
[0006] To solve the above technical problems, according to one aspect of the present invention, the present invention provides the following technical solutions: A human-machine collaborative control method for a heavy-load rescue robot, the steps are as follows: S1. Obtain the real-time speed of the mobile chassis of the heavy-load rescue robot , Current driving trajectory of the mobile chassis , current position, angular velocity of each joint of the rescue operation device , end effector spatial position , end effector load, and terrain elevation ,slope , obstacle height , obstacle distance ; S2. Based on the data collected in step S1, the state information and surrounding environment information of the heavy-duty rescue robot are evaluated, and the reward function of the safety reinforcement learning model is designed. The joint safety boundary of the mobile chassis and the rescue operation device during mobile operation in an emergency rescue environment is considered, and a safety reinforcement learning model based on a deep deterministic policy gradient algorithm is constructed to determine the control weights of the intelligent rescue operation device system. and intelligent chassis system control weight ; S3, the control signals of the operator, the intelligent rescue operation device system, and the intelligent chassis system are based on the control weights determined by the decision , Perform human-machine integration to obtain control signals of rescue operation devices and mobile chassis control signals , realizing human-machine collaborative control among the operator, mobile chassis and rescue operation device; , ; In the formula, is the weight of human-machine collaborative control of the rescue operation device, , They are the rescue operation device control signal and the operator's control signal of the operation device. is the human-machine collaborative control weight of the intelligent chassis system, , They are the control signals of the intelligent chassis control system and the control signals of the operator on the chassis system.
[0007] As a preferred solution of the human-machine collaborative control method for a heavy-load rescue robot described in the present invention, in step S1, the real-time speed of the mobile chassis of the heavy-load rescue robot is obtained by a speed sensor, a joint angular velocity sensor, an acceleration sensor, a position sensor, a force sensor, an inclination sensor and a laser radar. , Current driving trajectory of the mobile chassis , current position, angular velocity of each joint of the rescue operation device , end effector spatial position , end effector load, and terrain elevation ,slope , obstacle height , obstacle distance .
[0008] As a preferred solution of the human-machine collaborative control method for a heavy-load rescue robot described in the present invention, in step S2, the operation risk of the heavy-load rescue robot is analyzed by the data collected by the sensor. and driving risks Evaluation is performed in combination with the execution time of the operating device , Posture consistency between mobile chassis and working device and the matching degree between the operator's instructions and the heavy-duty rescue robot's execution feedback , design reward functions for secure reinforcement learning models; Reinforcement Learning Reward Function Defined as , are the gains of each parameter, The values of are all positive numbers.
[0009] As a preferred solution of the human-machine collaborative control method for a heavy-load rescue robot described in the present invention, the operation risk ; in, Represents the influence gain of joint angular velocity on operation risk, Indicates the impact gain of load on operational risk, It represents the influence gain of obstacles on operation risk. is the angular velocity of each joint, is the current load, It is the maximum load that the rescue operation device can bear. is the height of the obstacle, is the distance between the obstacle and the end effector.
[0010] As a preferred solution of the human-machine collaborative control method for a heavy-load rescue robot described in the present invention, the driving risk ; in, It represents the influence gain of real-time speed on driving risk. It represents the impact gain of obstacle collision information on driving risk. is the real-time speed of the moving chassis, is the maximum safe speed, is the height of the obstacle, is the distance between the obstacle and the mobile chassis, is the mass of the obstacle.
[0011] As a preferred solution of the human-machine collaborative control method for a heavy-load rescue robot described in the present invention, target consistency ; in, Indicates the real-time position of the mobile chassis. represents the location of the target object, Indicates the position of the end effector of the working device.
[0012] As a preferred solution of the human-machine collaborative control method for a heavy-load rescue robot described in the present invention, the matching degree between the operator's command and the heavy-load rescue robot's execution feedback is , , , ; in, are the gains of each parameter, and They are the operator's target attitude quaternion representation and the actual attitude quaternion representation, The maximum attitude error tolerated, Used to evaluate posture matching; is the speed of the operator's target command, is the actual speed of the heavy-load rescue robot, Acceptable speed difference, Used to evaluate speed matching; is the load expected by the operator, and is the actual load, Acceptable load difference, Used to evaluate load matching.
[0013] As a preferred solution of the human-machine collaborative control method for a heavy-load rescue robot described in the present invention, in step S2, the joint safety constraint boundary design steps of the safety reinforcement learning model are as follows: First, establish a global coordinate system, determine the positional relationship of each coordinate system, and construct a two-degree-of-freedom dynamic model of the mobile chassis considering lateral motion and yaw motion. The model is as follows:
[0014] in, , ; By discretizing the model, the following form is obtained: ; Where: is the lateral speed of the moving chassis, is the yaw rate, is the lateral displacement of the mobile chassis, is the yaw angle, is the front wheel cornering stiffness, is the rear wheel cornering stiffness, is the mass of the mobile chassis, is the longitudinal velocity of the moving chassis, is the distance from the front axle to the center of mass of the moving chassis, is the distance from the rear axle to the center of mass of the mobile chassis, is the moment of inertia of the moving chassis, is the front wheel turning angle, is the length of the discrete steps; The multi-degree-of-freedom model of the rescue operation device is further constructed, and the formula is as follows: , ; Where: is the position of the end effector of the rescue operation device in space, is the degree of freedom of the working device, It is The rotation angle of each joint, It's a joint The screw coordinates relative to the end effector coordinate system, is the terminal pose matrix at the initial position; The two-degree-of-freedom dynamics model of the chassis and the screw model of the rescue operation device are combined through the Jacobi transformation matrix to obtain a comprehensive joint model, which is expressed as follows:
[0015] In the formula, is the global Jacobian matrix of the heavy-duty rescue robot, is the Jacobian matrix of the mobile chassis, is the Jacobian matrix of the rescue operation device; Based on the Jacobian matrix of the joint model of the mobile chassis and the rescue operation device, environmental information such as obstacle distribution and distance, and state information such as chassis speed, joint angular velocity, and equipment size of the heavy-load rescue robot, a joint safety boundary model is constructed. The expression is as follows:
[0016]
[0017]
[0018]
[0019]
[0020] In the formula, is the tipping stability, which is a value between 0 and 1, where 0 means unstable and 1 means completely stable. is the ratio of the current chassis speed to the maximum speed is the ratio of the current joint angular velocity to the maximum angular velocity, is the collision avoidance distance, ensuring that the distance from obstacles is safe enough. The equipment size of the working device and chassis, considering its influence on stability, is the safety factor, which is determined by the static stability factor of the system , the dynamic stability factor of the system Environmental safety factors constitute, is the supporting moment, the moment of the chassis ground force on the tipping edge, is the tipping moment, the moment caused by the position of the center of gravity, is the maximum disturbance force that the system can withstand is the disturbance force currently applied to the system, is the coefficient of friction of the ground, is the normal support force of the ground on the system, is the total weight of the system, is the absolute value of the Jacobian matrix determinant of the joint model of the mobile chassis and rescue operation device; After obtaining the joint safety boundary, safety constraints are imposed on the state input consisting of the angular velocity of each joint, the spatial position of the end effector, the speed of the mobile chassis and the angle of rotation to limit the unsafe state, and finally the state information after safety constraints is output.
[0021] Compared with the prior art, the present invention has the following beneficial effects: the present invention dynamically and safely adjusts the human-machine collaboration weights between the operator and the intelligent chassis system, and between the operator and the intelligent rescue operation device system according to the real-time rescue environment information, the position information of the mobile chassis and the rescue operation device, and the joint safety constraint model, so as to realize the human-machine collaborative control of the three. Not only can the operation safety of the heavy-loaded rescue robot be effectively guaranteed through the joint safety constraint model, but also the operation efficiency of the heavy-loaded rescue robot in an emergency rescue environment can be improved through the efficient collaboration of the three. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the present invention will be described in detail below in combination with the accompanying drawings and detailed embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them: Figure 1 This is a flow chart of a human-machine collaborative control method for a heavy-load rescue robot according to the present invention; Figure 2 A human-machine collaborative control weight decision flow chart based on safety reinforcement learning provided by the present invention. DETAILED DESCRIPTION
[0023] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0024] A research route for a human-machine collaborative control method for a heavy-duty rescue robot Figure 1 First, the real-time speed of the mobile chassis of the heavy-duty rescue robot is obtained through speed sensors, joint angular velocity sensors, acceleration sensors, position sensors, force sensors, inclination sensors, laser radars and other devices. , Current driving trajectory of the mobile chassis , current position, angular velocity of each joint of the rescue operation device , end effector spatial position , end effector load, and terrain elevation ,slope , obstacle height , obstacle distance Secondly, based on the collected data, the status information and surrounding environment information of the heavy-duty rescue robot are evaluated, the reward function of the safety reinforcement learning model is designed, and the joint safety boundary of the mobile chassis and the rescue operation device during mobile operation in an emergency rescue environment is considered. A safety reinforcement learning model based on the deep deterministic policy gradient algorithm (DDPG) is constructed to determine the control weights of the intelligent rescue operation device system. and intelligent chassis system control weight Finally, the control signals of the operator, the intelligent rescue operation device system, and the intelligent chassis system are determined according to the control weights. , Perform human-machine integration to obtain control signals of rescue operation devices , mobile chassis control signal , realizing human-machine collaborative control among the operator, mobile chassis and rescue operation equipment.
[0025] Control signals of rescue operation equipment after human-machine integration : (1) Where: is the weight of human-machine collaborative control of the rescue operation device, , They are respectively the rescue operation device control signal and the operator's control signal of the operation device.
[0026] Mobile chassis control signals after human-machine integration : (2) Where: is the human-machine collaborative control weight of the intelligent chassis system, , They are the control signals of the intelligent chassis control system and the control signals of the operator on the chassis system.
[0027] Research routes for weight decision-making methods for human-machine collaborative control based on safety reinforcement learning include Figure 2 As shown, the present invention adopts a deep deterministic policy gradient algorithm (DDPG) with a continuous action space to train an intelligent safety decision-making model for a heavy-duty rescue robot, so as to realize dynamic optimization decision-making on the human-machine collaborative control weights of the rescue operation device and the mobile chassis. The DDPG algorithm has the ability to process continuous action space, so it can effectively adapt to the dynamic adjustment requirements of the control weights of the rescue operation device and the mobile chassis. This dynamic adjustment needs to balance the efficiency of human-machine collaboration and the safety of operation in a complex emergency rescue environment to ensure that the best decision can be made according to the real-time situation. In this method, in order to take into account the effectiveness of human-machine collaboration and the safety of the heavy-duty rescue robot operation, the design of the reward function and the joint safety constraint boundary becomes a key link. The operational risk of the heavy-duty rescue robot is measured by data collected by sensors. and driving risks Evaluation is performed in combination with the execution time of the operating device , Posture consistency between mobile chassis and working device , and the matching degree between the operator's instructions and the heavy-duty rescue robot's execution feedback , construct the reward function of the safety reinforcement learning model. At the same time, in the emergency rescue environment, the joint safety boundary of the mobile chassis and the rescue operation device is comprehensively considered, and the safety reinforcement learning method based on the DDPG algorithm is used to make a comprehensive optimization decision, so as to effectively realize the dynamic allocation of human-machine collaborative control weights and improve the safety of rescue operations and the efficiency of collaborative operations.
[0028] The present invention will focus on the construction of the reward function of the security reinforcement learning model and the design of the joint security constraint boundary of the security reinforcement learning model for detailed explanation below.
[0029] (1) Reward function construction of secure reinforcement learning model According to the acquired state information of the heavy-duty rescue robot, the reward function of the weight of intelligent decision-making human-machine collaborative control in safety reinforcement learning is reasonably designed. The present invention constructs a reward function for reinforcement learning based on the following aspects: Operational risk Driving risks , Execution time of the operation device , Posture consistency between mobile chassis and working device , and the matching degree between the operator's instructions and the heavy-duty rescue robot's execution feedback The design of this reward function comprehensively considers the operating risks of the rescue operation device and the driving risks of the mobile chassis system to ensure the safety of the robot during operation; at the same time, it combines the execution time of the operation device, the posture consistency between the mobile chassis and the operation device, and the matching degree between the operator's instructions and the robot's execution feedback to improve the efficiency of the robot's operation.
[0030] In an emergency rescue environment, the surrounding environment is complex and there are many obstacles. The operational risk of the rescue operation device is mainly affected by the joint angular velocity and the distribution of obstacles. Therefore, the operational risk for: (3) Where: Represents the influence gain of joint angular velocity on operation risk, Indicates the impact gain of load on operational risk, It represents the influence gain of obstacles on operation risk. is the angular velocity of each joint, is the current load, It is the maximum load that the rescue operation device can bear. is the height of the obstacle, is the distance between the obstacle and the end effector.
[0031] In addition to the operational risks faced by the rescue operation device, the driving risk of the mobile chassis system is also crucial. Since the driving of the mobile chassis is mainly affected by the distribution of obstacles, it is necessary to construct a driving risk for: (4) Where: It represents the influence gain of real-time speed on driving risk. It represents the impact gain of obstacle collision information on driving risk. is the real-time speed of the moving chassis, is the maximum safe speed, is the height of the obstacle, is the distance between the obstacle and the mobile chassis, is the mass of the obstacle.
[0032] During the joint operation of the mobile chassis and the working device, the movement postures of the two should serve the same target task, that is, they should have target consistency. For example, when performing a heavy object grasping task, the end effector of the working device should move toward the target object, and the mobile chassis should move to the location of the target object, so that the working device and the mobile chassis can efficiently and quickly complete the grasping operation of the target object. Based on this, when the target consistency between the mobile chassis and the working device is high, the efficiency of the collaborative control between the two will also increase. Therefore, the target consistency between the mobile chassis and the working device is used as part of the reward function, and the efficiency of the collaborative control of the joint operation is improved by maximizing the target consistency.
[0033] Constructing the consistency of the mobile chassis and rescue operation equipment for: (5) Where: Indicates the real-time position of the mobile chassis. represents the location of the target object, Represents the position of the end effector of the working device. This formula represents the rate of change per unit time of the sum of the squares of the position errors of the chassis and the end effector relative to the target object.
[0034] When the mobile chassis and the working device are consistent with the target, the square of the position error between their respective positions and the target object will gradually decrease. If the mobile chassis moves toward the target object, the working device also adjusts its posture to make the end effector approach the target object. Not only the square of the position error between the chassis and the end effector and the target object is reduced, but also because the working device is installed on the chassis, its posture adjustment further reduces the square of the position error between the end effector and the target object, thereby accelerating the reduction of the overall sum of square errors. Therefore, the unit time rate of change of the sum of square position errors of the mobile chassis and the end effector relative to the target object can significantly reflect the dynamic changes of target consistency.
[0035] In the coordinated control of the operator, mobile chassis and working device, the collaboration between the operator and the heavy-duty rescue robot is crucial. This is related to whether the robot can accurately execute the operator's instructions, thereby achieving efficient human-machine collaborative control. Therefore, it is of great significance to ensure that the action execution of the heavy-duty rescue robot is highly matched with the operator's control intention. By quantitatively analyzing the deviation between the target instructions issued by the operator and the actual execution results of the robot, the matching degree of the robot's action execution and the operator's control intention can be effectively described, thereby constructing the matching degree between the operator's instructions and the execution feedback of the heavy-duty rescue robot. for: (6) (7) (8) (9) Where: are the gains of each parameter, and They are the operator's target attitude quaternion representation and the actual attitude quaternion representation, The maximum attitude error tolerated, Used to evaluate posture matching; is the speed of the operator's target command, is the actual speed of the heavy-load rescue robot, Acceptable speed difference, Used to evaluate speed matching, is the load expected by the operator, and is the actual load, Acceptable load difference, Used to evaluate load matching.
[0036] By considering the operation risk and driving risk, the safety of the heavy-duty rescue robot is ensured; by considering the consistency of the posture of the mobile chassis and the working device, as well as the matching degree between the operator's instructions and the feedback of the heavy-duty rescue robot, the operating efficiency of the heavy-duty rescue robot under the coordinated control of the operator, the mobile chassis and the working device is improved. Based on the above set rules, the reinforcement learning reward function Defined as: (10) Where: are the gains of each parameter, The values of are all positive numbers.
[0037] (2) Joint safety constraint boundary design for secure reinforcement learning models Since the mobile chassis and the rescue operation device have a posture association during the mobile operation, it is necessary to comprehensively consider the state information of the two when establishing the joint safety boundary, so as to maximize the expansion of the operation space on the basis of ensuring the safety of the heavy-duty rescue robot. If the safety boundaries of the chassis and the operation device are considered separately, when only the state information of one party changes, the safety boundary of the other party will remain unchanged due to the lack of dynamic adjustment of the joint relationship between the two. However, the actual safety boundary will be reduced due to the consideration of the posture association between the chassis and the operation device. In this case, the setting of the independent safety boundary may not fully reflect the overall safety requirements of the system, thereby increasing the risk. For example, during the operation of the heavy-duty rescue robot, when the chassis turns in a narrow space, its movement will change the relative posture and operating range of the manipulator. At this time, if only the independent safety boundary of the chassis is considered, the situation that the manipulator enters the dangerous area due to the turning of the chassis may be ignored, thereby causing the manipulator to collide with surrounding obstacles. Therefore, it is very important to construct a joint safety constraint. The present invention constructs a joint model of the heavy-duty rescue robot based on the mobile chassis model and the operation device model, and designs a joint safety constraint of safety reinforcement learning based on this model.
[0038] First, establish the global coordinate system and determine the positional relationship of each coordinate system. Construct a two-degree-of-freedom dynamic model of the mobile chassis considering lateral motion and yaw motion. The model is as follows: (11) in . By discretizing the model described in (11), we can obtain the following form: (12) Where: is the lateral speed of the moving chassis, is the yaw rate, is the lateral displacement of the mobile chassis, is the yaw angle, is the front wheel cornering stiffness, is the rear wheel cornering stiffness, is the mass of the mobile chassis, is the longitudinal velocity of the moving chassis, is the distance from the front axle to the center of mass of the moving chassis, is the distance from the rear axle to the center of mass of the mobile chassis, is the moment of inertia of the moving chassis, is the front wheel turning angle, Is the distance walk length.
[0039] In order to accurately control the rescue operation device, its screw model was constructed, and the multi-degree-of-freedom model of the rescue operation device was further constructed. The formula is as follows: (13) (14) Where: is the position of the end effector of the rescue operation device in space, is the degree of freedom of the working device, It is The rotation angle of each joint, It's a joint The screw coordinates relative to the end effector coordinate system, is the terminal pose matrix at the initial position.
[0040] In order to realize the joint motion of the chassis and the working device, the two-degree-of-freedom dynamic model of the chassis and the screw model of the rescue working device are combined through the Jacobi transformation matrix to obtain a comprehensive joint model, which is expressed as follows: (15) Where: is the global Jacobian matrix of the heavy-duty rescue robot, is the Jacobian matrix of the mobile chassis, is the Jacobian matrix of the rescue operation device.
[0041] By real-time monitoring of the Jacobian matrix and its determinant of the joint model established above, the motion performance of the system is dynamically evaluated, and the safety range is adjusted according to the current state to prevent the system from entering a singular state, ensuring that the constructed safety boundary model fully reflects the motion coupling relationship between the mobile chassis and the rescue operation device, so that the joint safety boundary constraint is more effective and safer collaborative control is achieved. Therefore, based on the Jacobian matrix of the joint model of the mobile chassis and the rescue operation device, environmental information such as obstacle distribution and distance, and state information such as chassis speed, joint angular velocity, and heavy-duty rescue robot equipment size, a joint safety boundary model is constructed: (16) (17) (18) (19) (20) Where: is the tipping stability, which is a value between 0 and 1, where 0 means unstable and 1 means completely stable. It is the ratio of the current chassis speed to the maximum speed. It is the ratio of the current joint angular velocity to the maximum angular velocity. It is the collision avoidance distance, ensuring that the distance from obstacles is safe enough. It is the equipment size of the working device and chassis, considering its influence on stability. is the safety factor, which is determined by the static stability factor of the system , the dynamic stability factor of the system Environmental safety factors constitute. It is the supporting moment, the moment of the chassis ground force on the tipping edge. is the tipping moment, the moment caused by the position of the center of gravity. is the maximum disturbance force that the system can withstand is the disturbance force currently applied to the system. is the coefficient of friction of the ground. is the normal support force of the ground on the system. is the total weight of the system. It is the absolute value of the determinant of the Jacobian matrix of the joint model of the mobile chassis and rescue operation device.
[0042] After obtaining the joint safety boundary, safety constraints are imposed on the state input consisting of the angular velocity of each joint, the spatial position of the end effector, the speed of the mobile chassis and the angle of rotation to limit the unsafe state, and finally the state information after safety constraints is output.
[0043] Although the present invention has been described above with reference to the embodiments, various modifications may be made thereto and parts thereof may be replaced by equivalents without departing from the scope of the present invention. In particular, as long as there is no structural conflict, the various features in the embodiments disclosed in the present invention may be used in combination with each other in any manner, and the fact that these combinations are not exhaustively described in this specification is only for the sake of omitting space and saving resources. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.
Claims
1. A human-machine collaborative control method for a heavy-load rescue robot, characterized in that: Here are the steps: S1. Obtain the real-time speed of the mobile chassis of the heavy-load rescue robot , Current driving trajectory of the mobile chassis , current position, angular velocity of each joint of the rescue operation device , end effector spatial position , end effector load, and terrain elevation ,slope , obstacle height , obstacle distance ; S2. Based on the data collected in step S1, the state information and surrounding environment information of the heavy-duty rescue robot are evaluated, and the reward function of the safety reinforcement learning model is designed. The joint safety boundary of the mobile chassis and the rescue operation device during mobile operation in an emergency rescue environment is considered, and a safety reinforcement learning model based on a deep deterministic policy gradient algorithm is constructed to determine the control weights of the intelligent rescue operation device system. and intelligent chassis system control weight ; S3, the control signals of the operator, the intelligent rescue operation device system, and the intelligent chassis system are based on the control weights determined by the decision , Perform human-machine integration to obtain control signals of rescue operation devices and mobile chassis control signals , realizing human-machine collaborative control among the operator, mobile chassis and rescue operation device; ， ； In the formula, is the weight of human-machine collaborative control of the rescue operation device, , They are the rescue operation device control signal and the operator's control signal of the operation device. is the human-machine collaborative control weight of the intelligent chassis system, , They are the control signals of the intelligent chassis control system and the control signals of the operator on the chassis system.
2. The human-machine collaborative control method for a heavy-load rescue robot according to claim 1, characterized in that: In step S1, the real-time speed of the mobile chassis of the heavy-duty rescue robot is obtained through the speed sensor, joint angular velocity sensor, acceleration sensor, position sensor, force sensor, tilt sensor and laser radar. , Current driving trajectory of the mobile chassis , current position, angular velocity of each joint of the rescue operation device , end effector spatial position , end effector load, and terrain elevation ,slope , obstacle height , obstacle distance .
3. The human-machine collaborative control method for a heavy-load rescue robot according to claim 1, characterized in that: In step S2, the data collected by the sensor is used to determine the operational risk of the heavy-duty rescue robot. and driving risks Evaluation is performed in combination with the execution time of the operating device , Posture consistency between mobile chassis and working device and the matching degree between the operator's instructions and the heavy-duty rescue robot's execution feedback , design reward functions for secure reinforcement learning models; Reinforcement Learning Reward Function Defined as , ; are the gains of each parameter, The values of are all positive numbers.
4. The human-machine collaborative control method for a heavy-load rescue robot according to claim 3, characterized in that: Operational Risk ; in, Represents the influence gain of joint angular velocity on operation risk, Indicates the impact gain of load on operational risk, It represents the influence gain of obstacles on operation risk. is the angular velocity of each joint, is the current load, It is the maximum load that the rescue operation device can bear. is the height of the obstacle, is the distance between the obstacle and the end effector.
5. The human-machine collaborative control method for a heavy-load rescue robot according to claim 1, characterized in that: Driving risks ; in, It represents the influence gain of real-time speed on driving risk. It represents the impact gain of obstacle collision information on driving risk. is the real-time speed of the moving chassis, is the maximum safe speed, is the height of the obstacle, is the distance between the obstacle and the mobile chassis, is the mass of the obstacle.
6. The human-machine collaborative control method for a heavy-load rescue robot according to claim 1, characterized in that: Goal Alignment ; in, Indicates the real-time position of the mobile chassis. represents the location of the target object, Indicates the position of the end effector of the working device.
7. The human-machine collaborative control method for a heavy-load rescue robot according to claim 1, characterized in that: Matching degree between operator instructions and heavy-duty rescue robot execution feedback , , , ; in, are the gains of each parameter, and They are the operator's target attitude quaternion representation and the actual attitude quaternion representation, The maximum attitude error tolerated, Used to evaluate posture matching; is the speed of the operator's target command, is the actual speed of the heavy-load rescue robot, Acceptable speed difference, Used to evaluate speed matching; is the load expected by the operator, and is the actual load, Acceptable load difference, Used to evaluate load matching.
8. The human-machine collaborative control method for a heavy-load rescue robot according to claim 1, characterized in that: In step S2, the joint security constraint boundary design steps of the security reinforcement learning model are as follows: First, establish a global coordinate system, determine the positional relationship of each coordinate system, and construct a two-degree-of-freedom dynamic model of the mobile chassis considering lateral motion and yaw motion. The model is as follows: ； in, , ; By discretizing the model, the following form is obtained: ； Where: is the lateral speed of the moving chassis, is the yaw rate, is the lateral displacement of the mobile chassis, is the yaw angle, is the front wheel cornering stiffness, is the rear wheel cornering stiffness, is the mass of the mobile chassis, is the longitudinal velocity of the moving chassis, is the distance from the front axle to the center of mass of the moving chassis, is the distance from the rear axle to the center of mass of the mobile chassis, is the moment of inertia of the moving chassis, is the front wheel turning angle, is the length of the discrete steps; The multi-degree-of-freedom model of the rescue operation device is further constructed, and the formula is as follows: ， ； Where: is the position of the end effector of the rescue operation device in space, is the degree of freedom of the working device, It is The rotation angle of each joint, It's a joint The screw coordinates relative to the end effector coordinate system, is the terminal pose matrix at the initial position; The two-degree-of-freedom dynamics model of the chassis and the screw model of the rescue operation device are combined through the Jacobi transformation matrix to obtain a comprehensive joint model, which is expressed as follows: ； In the formula, is the global Jacobian matrix of the heavy-duty rescue robot, is the Jacobian matrix of the mobile chassis, is the Jacobian matrix of the rescue operation device; Based on the Jacobian matrix of the joint model of the mobile chassis and the rescue operation device, environmental information such as obstacle distribution and distance, and state information such as chassis speed, joint angular velocity, and equipment size of the heavy-load rescue robot, a joint safety boundary model is constructed. The expression is as follows: ； ； ； ； ； In the formula, is the tipping stability, which is a value between 0 and 1, where 0 means unstable and 1 means completely stable. is the ratio of the current chassis speed to the maximum speed is the ratio of the current joint angular velocity to the maximum angular velocity, is the collision avoidance distance, ensuring that the distance from obstacles is safe enough. The equipment size of the working device and chassis, considering its influence on stability, is the safety factor, which is determined by the static stability factor of the system , the dynamic stability factor of the system Environmental safety factors constitute, is the supporting moment, the moment of the chassis ground force on the tipping edge, is the tipping moment, the moment caused by the position of the center of gravity, is the maximum disturbance force that the system can withstand is the disturbance force currently applied to the system, is the coefficient of friction of the ground, is the normal support force of the ground on the system, is the total weight of the system, is the absolute value of the Jacobian matrix determinant of the joint model of the mobile chassis and rescue operation device; After obtaining the joint safety boundary, safety constraints are imposed on the state input consisting of the angular velocity of each joint, the spatial position of the end effector, the speed of the mobile chassis and the angle of rotation to limit the unsafe state, and finally the state information after safety constraints is output.