DRO-EDL-MPC: Evidential Deep Learning-Based Distributionally Robust Model Predictive Control for Safe Autonomous Driving
Hyeongchan Ham, Heejin Ahn
AI summary
Problem
Neural network perception uncertainties in autonomous driving compromise safety, yet existing robust control methods fail to adapt their conservativeness based on real-time model confidence or require impractical sample sizes.
Approach
The authors integrate Evidential Deep Learning into a distributionally robust optimization framework to dynamically size an ambiguity set based on perception confidence, which is then embedded into a computationally tractable Model Predictive Control algorithm.
Key results
- Proposes DR-EDL-CVaR, a safety constraint that constructs ambiguity sets using evidential distribution probabilities.
- Introduces DRO-EDL-MPC, a tractable motion planning algorithm that adapts conservativeness to perception confidence.
- Demonstrates in CARLA simulator that the method avoids collisions under high uncertainty while maintaining efficiency under high confidence.
- Provides a principled, sample-independent method for balancing robustness and performance in real-time robotic planning.
Why it matters
Enables safer, adaptive autonomous driving by bridging machine learning uncertainty quantification with real-time control, critical for deploying reliable self-driving systems.
Abstract
Safety is a critical concern in motion planning for autonomous vehicles. Modern autonomous vehicles rely on neural network-based perception, but making control decisions based on these inference results poses significant safety risks due to inherent uncertainties. To address this challenge, we present a distributionally robust optimization (DRO) framework that accounts for both aleatoric and epistemic perception uncertainties using evidential deep learning (EDL). Our approach introduces a novel ambiguity set formulation based on evidential distri- butions that dynamically adjusts the conservativeness according to perception confidence levels. We integrate this uncertainty- aware constraint into model predictive control (MPC), proposing the DRO-EDL-MPC algorithm with computational tractability for autonomous driving applications. Validation in the CARLA simulator demonstrates that our approach maintains efficiency under high perception confidence while enforcing conservative constraints under low confidence.