Path following of car-like vehicles using dynamic inversion

This paper focuses on a special path following task arising from the needs of vision-based autonomous guidance: a given front point of a car-like vehicle that is within the look-ahead range of a stereo vision system must follow a prespecified Cartesian path. A solution to this path following problem is provided by a new feedforward/feedback control strategy where the feedforward is determined by a dynamic generator based on exact dynamic inversion over the nominal vehicle model and the feedback is mainly issued by correcting terms proportional to the tangential and normal errors determined with respect to the vehicle's ideal trajectory. A convergence analysis of the resulting dynamic inversion based controller is established versus a vehicle's uncertain model defined via equation errors. Simulation examples highlighting the controller's performances are included.