Fast and Accurate Trajectory Generation for Non-Circular Omnidirectional Robots in Industrial Applications

This paper presents a new approach for fast and accurate generation of cubic spline trajectories for non-circular omnidirectional mobile robots and automated guided vehicles (AGVs), hereinafter referred to as mobile robots. This method can be combined with any existing path planner or guidance system for offline calculation of trajectories with independent translational and rotational movement. In addition, because of the capability to compute trajectories fast, this approach is suited for online generation of trajectories, in example for collision avoidance. The focus of this approach is on industrial applications, that is why accurate trajectories, accelerations and velocities of robots have to be predictable. For that reason, first with use of cubic splines a curvature continuous trajectory is generated and then velocity-time-profiles for translational and rotational movement with predefined maximum accelerations and velocities are calculated for each spline segment. Velocity-time-profiles are used to calculate poses and velocities for a closed-loop control, which generates the drive commands. In addition a Monte Carlo Particle Filter based localization is used to follow the planned trajectory. After a deviated pose of the planned trajectory is estimated by the localization, return-trajectories are calculated with quintic Bezier splines. They lead the robot back to the planned trajectory.