Analyzing and Improving Fault Tolerance of Learning-Based Navigation Systems
TimeThursday, December 9th2:30pm - 2:50pm PST
Event Type
Research Manuscript
Virtual Programs
Presented In-Person
Autonomous Systems
DescriptionLearning-based navigation systems are widely used in autonomous applications, e.g., robotics, unmanned vehicles and drones. Specialized hardware accelerators have been proposed for high-performance and energy-efficiency for such navigational tasks. However, transient and permanent faults are increasing in hardware systems and can catastrophically violate tasks safety. Meanwhile, traditional redundancy-based protection methods are challenging to deploy on resource-constrained edge applications. In this paper,we experimentally evaluate the resilience of navigation systems regarding algorithms, fault models and datatypes from both RL training and inference. We further propose two cost-efficient fault mitigation techniques that achieve 2x success-rate and 39% quality-of-flight improvements in learning-based navigation systems