Kinodynamic Trajectory Generation using Nonlinear Filters : a Multi-Dimensional Space Approach

The paper describes trajectory generation algorithms based on nonlinear smoothing filters, which can be applied to kinodynamic motion planning for mobile robots in a two-dimensional or three-dimensional space. The trajectory generators operate fully online to reach a given target point according to a pursuit-based logic. The logic sets reference inputs and dynamic constraints for a set of nonlinear filters, whose output signals are finally used to compute the trajectory compatibly with the kinematic model and the dynamic features of the robot tracking such a trajectory. Since the target point of the trajectory generator can be changed at any time during motion, the algorithm can execute online smoothing of straight-line reference paths, for example composed of viapoints assigned by a global planner on the basis of obstacle avoidance rules.