Longitudinal Control of an Autonomous Truck With Unobserved Gears

We provide a solution for the longitudinal control problem of heavy-duty vehicles when there is no information about the actual gear engaged by the automatic gearbox. The vehicle is modeled as a discrete-time Markov jump linear system based on a previous experimental identification procedure. The engaging of each gear corresponds to the operation modes in a Markov chain. We augment the longitudinal model, such that the information about the actual mode becomes an uncertain term. It is then possible to define an optimization problem whose solution yields a specific mode-independent regulator framework. Simulation results show that the obtained state feedback gain stabilizes the closed-loop system in spite of unobserved operation modes and adequately tracks the reference trajectories of states.