Efficient Neural Hybrid System Learning and Transition System Abstraction for Dynamical Systems
This work addresses the need for interpretable and verifiable dynamics learning in fields like control systems, though it appears incremental as it builds on existing neural and hybrid system methods.
The paper tackles the problem of learning and verifying dynamical systems by proposing a neural hybrid modeling framework that first trains a low-level model to approximate local dynamics and then abstracts it into a high-level transition system for computational tree logic verification, resulting in an interpretable and computationally efficient approach.
This paper proposes a neural network hybrid modeling framework for dynamics learning to promote an interpretable, computationally efficient way of dynamics learning and system identification. First, a low-level model will be trained to learn the system dynamics, which utilizes multiple simple neural networks to approximate the local dynamics generated from data-driven partitions. Then, based on the low-level model, a high-level model will be trained to abstract the low-level neural hybrid system model into a transition system that allows Computational Tree Logic Verification to promote the model's ability with human interaction and verification efficiency.