Most approaches to autonomous vehicle control are focused on normal driving conditions and are not suitable for on-limit control due to the non-linear nature of on-limit behaviour. Situations which require on-limit control are also commonly constrained by hard boundaries and require fast computation times, providing a unique set of problem constraints that are challenging to fulfill. Numerical simulation of on-limit manoeuvres often carries significant computational burden, precluding the direct use of full transient vehicle simulations in Model-Predictive-Control (MPC). Quasi-Steady-State (QSS) lap time simulation is a modelling method commonly found in motorsports, which uses point mass dynamics to control a vehicle within a set of predefined acceleration and jerk envelopes. QSS models are capable of accurately predicting the transient velocities and trajectories of on-limit vehicle motions, though vehicle controls such as steering and throttle cannot be formally tended to with this method. This paper presents a 3-tiered hierarchical control framework used to control autonomous vehicles in on-limit handling based on QSS motion. A simple constrained geometric path evaluation algorithm is used to determine a feasible path for initialisation, with QSS free-trajectory optimisation used to determine trajectory and velocity profiles. Finally, a short horizon transient vehicle model optimisation is used to convert QSS motion into a set of autonomous vehicle controls. The controller is evaluated in a Model-in-the-Loop setup, where the ability of the hierarchical framework to control the vehicle in near real-time is demonstrated at the limits of handling. Vehicle response is compared with a fixed horizon transient and fixed horizon QSS optimal control model, providing favourable results.
RMIT University: K. Tucker, R. Gover, R.N. Jazar, H. Marzbani