Coordinating Multiple Robots with Kinodynamic Constraints along Specified Paths

Jufeng Peng and Srinivas Akella

Abstract

This research focuses on the collision-free coordination of multiple robots with kinodynamic constraints along specified paths. The goal is to generate continuous velocity profiles for the robots. The approach consists of identifying collision segments along each robot's path, and then optimizing the motions of the robots along the collision and collision-free segments to avoid collisions and minimize the completion time. We optimize the robots motions by combining techniques from optimal control and mathematical programming. First, for each path segment for each robot, the minimum and maximum possible traversal times, while satisfying the dynamics constraints, are computed by solving the corresponding two-point boundary value problems. Using the times when a robot enters and exits a collision segment, the collision avoidance constraints for pairs of robots can be combined to formulate a mixed integer nonlinear programming (MINLP) problem. Since the resulting nonconvex MINLP model is difficult to solve, we describe two related mixed integer linear programming (MILP) formulations that provide schedules that are lower and upper bounds on the optimum. The approach is illustrated with coordination of multiple robots, modeled as double integrators subject to velocity and acceleration constraints. Results illustrating coordination of 12 robots are described. We also describe the application of this approach to nonholonomic car-like robots.

Animations

Note:

When two robots collide, their color changes to red.
The gif animations repeat in a loop. The animations may sometimes appear a bit jerky.
The avi movies may be best viewed with Windows Media player. They use the DivX codec , which you may need to install.



Srinivas Akella / sakella@cs.rpi.edu