Path following for autonomous vehicle navigation based on kinodynamic control

This paper addresses the problem of path following for mobile robots with particular emphasis on integrating the global path planning, path following and a collision avoidance scheme in a unified framework. Whereas the traditional path following algorithms aim at minimizing an error function with respect to a given path and kinematic and/or dynamic model of the robot, the problem of collision avoidance is often neglected or simply cast to the replanning phase of the global planner which issues the given path. Such approaches that do not check explicitly for collision for the given state of the ego-robot and the environment can easily lead to hazardous situations, in particular if latencies are present in the global path planning phase. In order to address obstacle avoidance directly, a navigation framework is presented here that combines a path following control scheme to attain a global objective with a collision checking scheme that incrementally builds collision-free trajectories, thus ensuring ego-robot safety at all times, with respect to the partially known static environment obstacles and kinodynamic limitations of the ego-robot itself. Two novel path following schemes are presented, namely the Traversability-anchored Dynamic Path Following (TADPF) and a combined TADPF-Sliding Mode Path Following (SMPF), based on a previously developed SMPF technique. The two path following schemes have been verified both in simulation and experimentally on a test Ackermann-like vehicle.