Collision-free trajectory planning for multiple robots is considered. The multiple robot systems handled in the dissertation are divided into three groups according to the characteristics of their applications : i) the case when only the starting and final positions of the robots are specified, ii) the case when the paths are given, and iii) the case when the paths and the coordination are specified.
For the case i), the configurations of robots are taken as the variables of the neural circuit, and the energy of network is determined by combining various functions, in which one function is to make each robot approach to its goal and another helps each robot form colliding with other robots. Also a differential equation of the circuit which tends to minimize the energy is derived.
For the case ii), the robot dynamics is transformed as a function of the traveled lengths along the path, and the bounds on acceleration and velocity are described in the phase plane by taking the constraints on torques and joint velocities into consideration. Collision avoidance and time optimality are considered simultaneously in the coordination space and the phase plane, respectively.
For the case iii), by describing the robot dynamics as a function of the traveled length along a coordination curve, the two-robot system is transformed to a virtual one-robot system, and a method for minimum-time velocity planning is derived in the phase plane and converted into a form to which simple iterative calculation can be applied.
Finally in order to make the robots faithfully follow the planned path, structure of a controller with learning capability through the characteristics of neural network is proposed and discussed.