Enhancing Fault Tolerance and Resilience in Autonomous Ground Vehicles through Detection and Diagnosis of Physical and Digital Attacks on Perception Systems

Project begins 2025.

As autonomous ground vehicles (AGVs) are increasingly expected to operate in contested environments, they face deliberate physical and cyber attacks aimed at degrading their perception systems. To maintain operational effectiveness, both automated systems and human operators must detect, diagnose, and respond to such attacks in realtime. Threat alerts, vehicle status, and available countermeasures need to be presented to human operators in a manner that ensures fast and accurate responses without overloading the operator.

This project aims to develop effective user interfaces and threat detection models to support robust fault detection and fault-tolerance techniques for AGV perception systems under attack. The focus is on supporting detection and mitigation of the effects of both physical and cyber disruptions, ensuring resilient operation in hostile and unpredictable environments.

The overall research objective is to investigate methods for maintaining safe and effective operations in the face of ongoing physical and cyber attacks on an AGV by supporting detection of and response to the attacks. The research effort will achieve this objective by (1) characterizing physical and cyber threats to the AGV perception systems and downstream components (e.g.,occupancy grid generation and control algorithms), (2) designing and evaluating user interfaces that enable operators to make rapid, informed decisions based on real-time threat detection, and (3) developing AI/ML models and associated methods for detecting, diagnosing, and responding to threats.

The project will leverage an existing cyber-physical systems simulation environment to develop human in-the-loop simulations of physical and cyber attacks on AGV systems to support evaluation of the effectiveness of operator responses and the AI/ML detection mechanisms.

Previous related ARC projects:

• Evaluating Sensitivity of Autonomous Algorithms to Sensor Error and Environmental Conditions
• Recognizing and Reconstructing Distorted and Obscured Perceptual Sensor Data Resulting from Soiling of the Sensor

#1.43