EnergyShield: Provably-Safe Offloading of Neural Network Controllers for Energy Efficiency

TL;DR

EnergyShield enables safe offloading of neural network controllers in autonomous driving by providing formal safety guarantees and energy savings through a runtime safety monitor, validated in simulation.

Contribution

It introduces a formal, provably-safe framework for offloading NN controllers to edge computing, ensuring safety and energy efficiency in autonomous driving.

Findings

Achieves 24-54% energy savings compared to on-vehicle NN evaluation.

Maintains vehicle safety under various wireless and delay conditions.

Provides a formal quantification of tolerable edge response time.

Abstract

To mitigate the high energy demand of Neural Network (NN) based Autonomous Driving Systems (ADSs), we consider the problem of offloading NN controllers from the ADS to nearby edge-computing infrastructure, but in such a way that formal vehicle safety properties are guaranteed. In particular, we propose the EnergyShield framework, which repurposes a controller ''shield'' as a low-power runtime safety monitor for the ADS vehicle. Specifically, the shield in EnergyShield provides not only safety interventions but also a formal, state-based quantification of the tolerable edge response time before vehicle safety is compromised. Using EnergyShield, an ADS can then save energy by wirelessly offloading NN computations to edge computers, while still maintaining a formal guarantee of safety until it receives a response (on-vehicle hardware provides a just-in-time fail safe). To validate the benefits of EnergyShield, we implemented and tested it in the Carla simulation environment. Our results show that EnergyShield maintains safe vehicle operation while providing significant energy savings compared to on-vehicle NN evaluation: from 24% to 54% less energy across a range of wireless conditions and edge delays.