Learning-enabled Parameter Synthesis for Nonlinear Systems from Signal Temporal Logic
It addresses the problem of safe parameter synthesis for nonlinear control systems with formal guarantees, which is important for safety-critical applications.
This work proposes a gradient-based optimization method combined with set-based reachability verification to synthesize parameters for nonlinear systems that robustly satisfy continuous-time STL specifications under uncertain initial conditions, achieving provable guarantees and scaling to 18 parameter dimensions.
Signal Temporal Logic (STL) is increasingly used to describe interpretable objectives and constraints for optimal control and learning methods, especially when no target time series data is available. In this work, we propose to synthesize parameters for nonlinear systems that robustly satisfy continuous-time STL specifications for uncertain initial conditions. To this end, we use gradient-based optimization along with set-based reachability verification to efficiently learn in high-dimensional parameter spaces while providing provable satisfaction guarantees for the optimized parameters. We demonstrate the effectiveness and scalability of our method on three systems with up to 18 parameter dimensions.