Research On Force-tactile Perception For Physical Interaction Control Between Robot And Dynamic Environment

For future open operation missions,robots will be more exposed to uncertain environments with dynamic effects.It is a challenge for complex and dexterous operations that have to meet both force/dynamics and position requirements to ensure operational performance.Furthermore,it is highly costly and time-consuming to guide by vision under a narrow and obscured uncertain environment with dynamic effects specifically.To address the above problems,this paper analyzed the coupled dynamics influencing robotic end operations under uncertain environments with dynamic effects,and proposed a control framework for robotic autonomy under dynamic uncertain environments inspired by human tactile exploration behavior to achieve safety and autonomy in operational missions with force and position requirements under visually obstructed conditions.Firstly,based on the analysis of the typical operating environment,an equivalent dynamic environment model and the interaction model between the robot and the environment were established.At the same time,considered the effect of joint clearances on the interaction characteristics of robot and environment,a joint clearance model with solid lubrication was introduced,and then the whole system model was established.Secondly,the control problems that exist when the robot interacts with the dynamic environment were analyzed with the example of on-orbit construction,and a force-tactile perception strategy was designed.In this strategy,a probabilistic force-tactile exploration method using Bayesian networks,which was achieved by a hybrid Monte Carlo sampling algorithm and particle filtering,was proposed to perceive uncertain environmental features.And based on the fuzzy plain Bayesian principle,the memory adjustment method and inertial thinking method were introduced to correct the target position and shape features of the environment,respectively.The fusion decision strategy was designed by fusing the probabilities of force and position based on the D-S theory,which guided the robot motion after each acquisition of the estimated environmental features.It also enabled the robot to judge whether the desired operation target has been achieved or the feature estimate needs to be updated.Then,to ensure the safety of the operating system,the pliability model was introduced into repeatedly perform exploration,planning and execution steps to reduce interaction forces and the number of exploration thus optimizing the overall control framework.A two-loop robotic autonomy control framework was proposed,where the outer loop handled the robotic interaction behavior with the dynamic environment by combining a force-tactile perception strategy and a pliability model.In the inner loop,the robotic joint motion control is implemented to further optimize the system in terms of motion stability by compensating the unknown part of the robot model with a sliding mode term and reducing the impact of the robot model error and joint transmission with current compensation based on neural networks.Finally,a simulation system for on-orbit construction missions and a ground experiment platform with dynamic effect simulation of the operating environment were established.The robot independently completed the construction missions under the conditions of incomplete visual information.Simulation and experiment results jointly validated the effectiveness of the framework in terms of intelligence,autonomy,and safety.