Real-time trajectory generation for car-like vehicles navigating dynamic environments

This paper presents Tiji, a trajectory generation scheme, ie an algorithm that computes a feasible trajectory between a start and a goal state, for a given robotic system. Tiji is geared towards complex dynamic systems subject to differential constraints, such as wheeled vehicles, and its efficiency warrants it can be used in real-time. Above all, Tiji is able to compute a trajectory that reaches the goal state at a prescribed final time in order to avoid collision with the moving objects of the environment. The method proposed, which relies upon a parametric trajectory representation, is variational in nature. The trajectory parameters are incrementally updated in order to optimize of a cost function involving the distance between the end of the trajectory computed and the (goal state, final time) pair. Should the goal state be unreachable (if the final time is ill-chosen), the method returns a trajectory that ends as close as possible to the (goal state, final time) pair, which can be useful in certain applications.