This work presents a technique for the kinodynamically feasible trajectory planning for uncertain vehicles exploiting pattern recognition. A kinodynamically feasible trajectory consists into a path and a proper timing law which abides by given kinematic and dynamic constraints, which are of capital importance for the safe motion of the vehicle, in particular in shared environments. In the proposed approach planning is performed on a graph whose nodes represents the operationalscenario. Candidate trajectories are generated through a low-complexity approach (rapidly exploring random trees) and those with the highest probability of being kinodynamically feasible are subject to a set-based verification to guarantee their feasibility. The proposed approach is validated through simulations in the Simulink/MATLAB environment involving a vehicle.
Nardi et al. (Fri,) studied this question.