Frameworks for interfacing trajectory tracking with predictive trajectory guidance for autonomous road vehicles

This paper investigates two frameworks for interfacing trajectory tracking functions with a computationally tractable nonlinear model predictive trajectory guidance module for an autonomous road vehicle. In the first framework, the predictive trajectory guidance itself is configured in a tracking mode where the control inputs computed by the model predictive control act as targets for some lower-level control system acting on the steering, brakes or engine. In the second framework, the state trajectories computed by the predictive trajectory guidance module are configured to be tracked by a non-predictive state tracking controller derived via input/output linearization. In both frameworks, the main considerations are driven by the time scales selected for computational expediency of the predictive guidance module, its update rate, and for control of the lower-level dynamics. The performance of the two frameworks, including the computational aspects, are compared considering public road driving as well as high performance race line scenarios.