Paper Accepted at IEEE Control Systems Letters!
Excited to announce that our paper “Verifiable Safety Q-Filters via Hamilton-Jacobi Reachability and Multiplicative Q-Networks” has been accepted for publication in IEEE Control Systems Letters (L-CSS)!
This work addresses a critical challenge in safe reinforcement learning: how to guarantee safety constraints while learning optimal control policies. Our approach combines:
- Hamilton-Jacobi reachability analysis for formal safety verification
- Multiplicative Q-networks for efficient policy learning
- Real-time safety filtering that can be deployed on robotic systems
The method provides provable safety guarantees while maintaining high performance in complex control tasks.
Many thanks to my co-authors Hanjiang Hu, Yujie Yang, and Prof. Changliu Liu for their collaboration on this project!