Research On Key Technology Of Nonholonomic Mobile Robot Motion Planning

With the development of society and more severe population aging phenomenon,as well as the continuous improvement of human cost,the demand for intelligent mobile robots is increasingly urgent.Motion planning,one of the key technologies to realize robot intelligence,attracts wide attention of academia and industry.In most situations of motion planning,the robot is considered as an unconstrained mass point.Consequently,the motion planning problem can be regarded as finding a collision-free sequence of valid configurations when the robot moves from the source to destination.This method is extensively used in engineering for its simplicity in constructing and solving.However,most wheeled mobile robots suffer from nonholonomic constraints,which cause that the robot can not follow any trajectory.In the case of high-speed movements and/or heavy load transportation,it will be hard to avoid the obstacles in time and follow the trajectory perfectly without the consideration of nonholonomic constraints.Therefore,it is very important to consider the nonholonomic constraints in the process of motion planning.In order to avoid the occurrence of above contradictions it is necessary to generate a trajectory complying with the natural motions of a mechanical system.In this dissertation,the mechanism of nonholonomic constraints and its influence on motion planning are analyzed,and the motion planning for nonholonomic mobile robot in the environment with static obstacles and dynamic obstacles is considered.Major researches are summarized as follows:1.The dynamical modeling of the nonholonomic differential-drive mobile robot(DDMR)is considered.Based on Lagrangian mechanics,the general dynamic model of DDMR is developed with the center of mass point as the reference point of generalized coordinate,where the Lagrange multipliers are introduced to solve the problem of nonholonomic constraints.In addition,the controllability of DDMR planning among obstacles is proven,which provides theoretic basis for nonholonomic constrained motion planning problem.In the field of motion planning,the dynamic model based on the center of mass point is better than mid-point on the axis,because it will supply more free space.2.A Multiple-Interval Chebyshev Pseudospectral(PS)method for solving the global motion planning problem of nonholonomic DDMR is presented.The time interval is divided into several sub-intervals based on multiple-interval strategy,and the state and control variables of particular nodes are expanded in the Chebyshev polynomial of order N.Then,the other state and control variables are approximated by the interpolation method.Hence,the infinite dimensional motion planning problem is converted into a finite-dimensional nonlinear programming problem where is easy to be solved.The multiple-interval strategy can deal with the problem that the interpolation of trajectory may violate the obstacles caused by the lack of enough nodes nearby the obstacles.3.A novel priority-based Non-Cooperative Distributed Model Predictive Control(PBNC-DMPC)method is presented based on Net-MPC to solve the multiple-robot motion planning problem.The urgency value of collision avoidance is quantized based on a novel strategy of priorities assignment.Consequently,the priorities generate a partial order which is a valid sequence for the agents to solve their optimization problems.The problem of prediction consistency caused by the time-variant coupling topology in multiple-robot motion planning and the conflicts in their constraints caused by an inappropriate priorities assignment can be solved by this method.4.A new method based on optimal reciprocal collision avoidance(ORCA)for solving coordinative navigation of multi-nonholonomic-robots is presented.The optimal velocity of collision avoidance and discrete ”safe zones” are delivered in each ORCA step.With the proposed algorithm,the optimal control input can be obtained based on the rolling optimization property of model predictive control,and the situation,where the nonholonomic constraints can not be handled in velocity obstacle space,can be addressed up.This method does not rely on the communication among robots.Naturally,it is regarded as a suitable methodology for dealing with the coordinative motion planning of large-scale multiple-robots.In this dissertation,the influence of nonholonomic to motion planning is deeply researched and the difficulty points of motion planning with the consideration of nonholonomic constraints are detailedly analyzed.On the basis of existed theories and technologies,the corresponding algorithms,which can facilitate the development of autonomous navigation technology and theory,are proposed.