We use nonlinear model predictive control (MPC) for autonomous trajectory planning and guidance of ground vehicles in dangerous driving situations. Nonlinear moving horizon estimation (MHE) is used to observe the vehicle's state driving conditions. we exemplarily present an obstacle avoidance scenario. A vehicle is following a reference line (e.g., a road) in difficult drivin conditions (e.g., ice- or snow-covered road) automatically avoiding obstacles that pop up. A trajectory of a vehicle travelling on that road could then look like this:
A vehicle avoiding obstacles on an icy road. The figure is taken from [1].
To achieve this performance we use a combined nonlinear MHE/MPC scheme on basis of a 12 state rigid body dynamics vehicle model, including position, orientation, directional and rotational velocities in the x-y plane, wheel dynamics, and load transfer. The road friction coefficient is estimated online alongside with the states. Real-time feasible computations are obtained using an efficient auto-generated solver through the ACADO Code Generation tool. A list of paramters used for simulation can be found here.
References:
[1] Janick V. Frasch, Andrew J. Gray, Mario Zanon, Hans Joachim Ferreau, Sebastian Sager, Francesco Borrelli, and Moritz Diehl. An Auto-generated Nonlinear MPC Algorithm for Real-Time Obstacle Avoidance of Ground Vehicles. European Control Conference 2013.
[2] Mario Zanon, Janick V. Frasch and Moritz Diehl. Nonlinear Moving Horizon Estimation for Combined State and Friction Coefficient Estimation in Autonomous Driving. European Control Conference 2013.