Study On Robot Compliant Control Strategy And Motion Planning

In order to cope with complex operating mission,collaborative robots not only need to have strong controllability and flexibility,but also need to have the ability to finish complex work autonomously and intelligently.However,current robot control systems face the problems such as low efficiency of motion programming and insufficient control flexibility,etc.In addition,the shortcomings of traditional compliant control methods which mainly rely on sensors and environmental models have been difficult to complete the increasingly complex tasks.Therefore,how to improve the skills learning,generalization and interaction capabilities of robotic systems has become a key step to stimulate the development of next generation collaborative robots.On one hand,this thesis designs a robotic incremental motion controller,focusing on tackling the problems of difficult to determine parameters of the neural nodes and single applicability in traditional neural network controller such that improving the skills learning and generalizing ability in robot motion control.On the other hand,aiming at solving the difficulty in obtaining the impedance model during the interaction between robot and environment,the compliant motion control under unknown environment is studied and an efficient and safe robot-environment interaction technology is developed.To be specific,the main contributions of this thesis include three aspects as follows.(1)Efficient human-like motion control and skills learning ability are very important for intelligent robots to achieve complex tasks.Based on broad neural network,this thesis proposes a novel incremental learning motion control framework.Thanks to the outstanding incremental learning and generalizing capability of broad neural network,we first design an incremental motion controller for different task trajectories which is integrated with barrier Lyapunov function and error constraint functions.However,the size of neural network compact set is difficult to determine since the number of neural nodes in broad learning system is dynamic.By designing a switching function and virtual dynamic neural network compact set in this thesis,we can ensure that the overall broad neural network controller is ultimately bounded and stable such that the global stable incremental motion control can be achieved.Finally,the incremental motion controller can be obtained through learning different tasks incrementally and the neural network controller whose weights have converged can be reused to new task by introducing the deterministic learning theory and persistent excitation condition.(2)Advanced adaptive impedance control technology is the prerequisite for friendly and safe robot-environment interaction.When robot interacts with the environment,the expected task trajectory will be changed due to the external disturbance,and the performance of the original controller may be affected.In addition,the actual environment model is difficult to obtain,and achieving compliant control in unknown environments is a challenging subject.This thesis designs a novel incremental fuzzy neural network based state constraint motion controller.First,the optimal impedance gain can be obtained and optimal impedance control is achieved by designing the cost function of the interaction between the robot and the environment,and solving the optimal model of the environment through the policy iterative learning method.Meanwhile,the control jitter problem which is caused by the sudden change of impedance parameter can be addressed by designing a soft function.Finally,the parameters of feature nodes in broad fuzzy neural network can be adjusted adaptively in terms of motion trajectory and robot state.In this case,the robot is able to work in the safe area by designing a reasonable barrier Lyapunov function in order to ensure the transient and steady-state tracking performance of the robot.(3)Impedance control is suitable for robot-environment compliant interactions.However,because it is difficult to obtain the impedance model of the environment,accurate force-position control is notoriously difficult to achieve.In this regard,this thesis proposes a novel trajectory adaptive compliant interaction algorithm based on dynamic system theory.In order to achieve safe interaction under unknown environment and tackle the problem of relying on reasonable environmental impedance model for traditional compliance control,a potential function and robust factor is established in terms of force feedback and force tracking error to adjust the expected trajectory derived from the dynamic system adaptively which is able to ensure the stability and convergency of adaptive trajectory,so as to perform a compliant interaction between robot and environment under unknown environment.Thereby,the problem of traditional compliant control dependent environmental impedance model is solved from the perspective of motion planning.