Adversarial Reinforcement Learning for Observer Design in Autonomous Systems under Cyber Attacks
This addresses sensor and communication failures in autonomous control systems, but it appears incremental as it applies existing adversarial methods to a specific domain.
The paper tackles the problem of designing robust observers for autonomous systems under cyber attacks by developing an adversarial deep reinforcement learning framework, and simulation results show the learned strategies perform well when adversarial errors are bounded.
Complex autonomous control systems are subjected to sensor failures, cyber-attacks, sensor noise, communication channel failures, etc. that introduce errors in the measurements. The corrupted information, if used for making decisions, can lead to degraded performance. We develop a framework for using adversarial deep reinforcement learning to design observer strategies that are robust to adversarial errors in information channels. We further show through simulation studies that the learned observation strategies perform remarkably well when the adversary's injected errors are bounded in some sense. We use neural network as function approximator in our studies with the understanding that any other suitable function approximating class can be used within our framework.