Our team member Baha Zarrouki represented our AVS lab at this years American Control Conference (ACC24) in Toronto, Canada. Baha showcased our recent work on model predictive control algorithm for autonomous driving trajectory following. Especially here, a Stochastic Nonlinear Model Predictive Controller (SNMPC) twas presentend that work on a real-time applications in real-world scenarios. SNMPC's effectiveness in real-time control of a high-dimensional, highly nonlinear system: the combined longitudinal and lateral motion control of an autonomous passenger vehicle.

🌟 What makes our algorithm special?
The SNMPC leverages probabilistic uncertainty descriptions, allowing the incorporation of acceptable risk levels in system operation. This approach strikes a balance between closed-loop performance and constraint violations, mitigating the inherent conservativeness associated with Robust Nonlinear MPC.
We introduced the Uncertainty Propagation Horizon (UPH) concept to limit the time for uncertainty propagation through nonlinear system dynamics. This prevents divergence in the evolution of uncertain states, avoids overly tightened constraints, leverages feedback loop advantages, and reduces computational overhead. UPH effectively addresses infeasibility issues, even with incorrect uncertainty assumptions or strong disturbances.

The SNMP was validated with TUM-CONTROL in simulation and in a real world environment. You can check out the paper here: 
📄 "A Stochastic Nonlinear Model Predictive Control with an Uncertainty Propagation Horizon for Autonomous Vehicle Motion Control"
👨‍🔬🙏 Special thanks to co-authors Chenyang Wang and Johannes Betz

🔗 Link: arxiv.org/abs/2310.18753 (initial version)