Motion Planning Of Spherical Robot Based On Particle Swarm Optimization Algorithm

Spherical robots are self-governed mobile with a global or approximately global shell, and take the rolling as their main motion mode. The spherical robots have the advantages such as: obtaining the great steady ability after being imposed a disturbance, achieving the omnidirectional rolling and agile swerving compared with wheeled robots and tracked robots and humanoid robots, because of the special shell and motion mode. At present, spherical robots are still studied in laboratories. There are many factors limiting the applications of robots in practical fields. However, the facus is that there is not a good motion planning for spherical robots.The spherical robot to be researched has a stable platform and can move in all directions. In this paper, the motion planning is detailedly studied.Firstly, a newly emerged efficient evolutionary computation technique-the Particle swarm optimization algorithm is investigated, including the principal theory, methods to realization and current application. Aiming at the problem that the particle swarm optimization (PSO) is difficult to deal with premature and localconvergence, an improved particle swarm optimization based on feedback mechanism was introduced. Simulation results show that the improved PSO not only effectivelyalleviates the problem of premature convergence, but also has fast convergence speed.The nonholonomy and the nonlinear controllability for a nonholonomic spherical robot are discussed firstly. The kinematic differential equations of the spherical robot are changed into an affine nonlinear control system without drift. It is indicated that the spherical robot is a completele nonholonomic system because the distribution spanned by control vector fields is not involutive. It is proved that spherical robot is a nonlinear controllable system according to the controllability rank condition of Lie algebra. Then the optimal control problem of nonholonomic motion planning of the spherical robot is discussed. Based on the optimal control principle and techniques of optimum, a numerical algorithm of continuous-control motion planning is proposed based on particle swarm optimization algorithm. And two different inputs aer used to control the motors. For a number of missions, it is shown that the method agrees well with the result, and the quadratic polynomial has the better effect.