On-the-Fly Symbolic Synthesis With Memory Reduction Guarantees

Abstract

We present novel on-the-fly abstraction-based controller synthesis algorithms for two fundamental specifications—invariance and optimal reachability. In contrast to existing methods, the algorithms we propose do not require precomputation or storage of any large abstraction parts. This translates to orders of magnitude of reduction in memory consumption compared to standard methods. Moreover, our algorithms fully apply to general nonlinear systems and provide controllers that do not posses any additional conservatism. We demonstrate the successful synthesis of controllers for a problem that was previously unsolvable due to excessive memory and time consumption.