Robotic Natural Motion Planning For Human-robot Interaction

Because of the increasingly complex task of the robot and changing requirements,how to make a robot with flexible configuration and motion planning capabilities to meet the needs of complex tasks,has become the core issue of the current motion planning field research.In practical interaction process,in order to complete the robot with a human user of natural and effective interaction,the robot needs to appropriately plan of their own movement,and make the movement pattern similar to humans based on mission requirements and constraints.Therefore,how to give the robot the ability to plan a natural movement,and then apply to the complex task of human-computer interaction,has important academic value and application significance.In this thesis,how to efficiently and effectively meet the mission requirements of natural motions under the background of human-machine interaction is the core topic.A representation space based robotic natural motion planning algorithm is designed with both feasibility and efficiency considered,and the simulation results show the effectiveness of the algorithm.The main work is summarized as follows:1)An interactive motion planning framework was proposed based on representation space,in order to achieve an iteration and interaction between task requirements and motion.As the existing robot motion planning algorithms could not adapt to the needs of complex tasks,the idea of interaction was introduced to motion planning progress,which leads to an interactive motion planning framework.This framework generates dynamic iteration between the task layer and motion layer for complicated task constraints.2)Task-oriented natural motion constraints are designed to integrate the path generated by the robot motion planner with the task execution mode of human.As the existing motion planning algorithms seldom consider whether the robot is similar to humans when executing a same interaction task,the robot may generate inconsistent motion comparing with humans when performing interaction tasks.Thus,in this thesis,natural motion constraints are designed for specific tasks to make a robot complete the task closer to humans.3)A multi-scale space decomposition method was designed to establish the connection between each reachable working subspace,and effectively understand the topology of the whole working space.Due to the natural motion constraints,the dimension of representation as well as the computation to solve planning tasks will increase.This thesis presents a multi-scale decomposition method to depict the topology of the working space on a higher level,which can efficiently cover most reachable space of the robot.4)A multi-scale neighborhood based rapidly-exploring random tree algorithm was proposed to significantly reduce the range of working space to be explored,so as to enhance the efficiency of solving planning tasks.In order to avoid the planner exploring unreachable and non-optimal subspace,the space to be explore is restricted by the proposed algorithm on the level of multi-scale neighborhoods,so that the planner can be concentrated in the subspace which most likely contains the optimal path.