Safe Control for High-Speed Driving using Decomposition-Based Reachability Analysis

Sammanfattning

This thesis presents a computationally efficient approach for determining safe control input ranges in high-speed driving by leveraging reachability analysis with system decomposition. High-speed driving scenarios call for high-order system models that can render the computation of safety guarantees intractable. Recently, decompositionbased methods have emerged to address such challenges. Building on this work, the thesis applies decomposition-based reachability analysis to a practical scenario involving a 7D control-affine single-track model, derived from the 1:10 car SVEA research platform. To extend the utility, a novel method is proposed to compute safe control input ranges directly from the decomposed system outputs, minimizing memory usage while maintaining performance. Additionally, several challenges associated with the implementation of decomposition-based reachability analysis are discussed, including complexities with high-order models and the selection of system decomposition. The proposed method is evaluated in a simulated high-speed driving scenario where a vehicle must navigate safely through a narrow gap. The results demonstrate that decomposition-based reachability analysis can provide rich insights into scenario safety that extend beyond the assessed scenario and promise great utility in system design and control.