GEMINUS: Dual-aware Global and Scene-Adaptive Mixture-of-Experts for End-to-End Autonomous Driving

TL;DR

GEMINUS introduces a dual-aware mixture-of-experts framework for end-to-end autonomous driving, combining global robustness with scene-specific adaptability, leading to superior performance in diverse traffic scenarios.

Contribution

The paper presents GEMINUS, a novel mixture-of-experts architecture with a dual-aware router for dynamic expert activation, enhancing adaptability and robustness in autonomous driving.

Findings

Outperforms existing methods in Bench2Drive benchmark

Achieves state-of-the-art Driving Score and Success Rate

Operates effectively with monocular vision input

Abstract

End-to-end autonomous driving requires adaptive and robust handling of complex and diverse traffic environments. However, prevalent single-mode planning methods attempt to learn an overall policy while struggling to acquire diversified driving skills to handle diverse scenarios. Therefore, this paper proposes GEMINUS, a Mixture-of-Experts end-to-end autonomous driving framework featuring a Global Expert and a Scene-Adaptive Experts Group, equipped with a Dual-aware Router. Specifically, the Global Expert is trained on the overall dataset, possessing robust performance. The Scene-Adaptive Experts are trained on corresponding scene subsets, achieving adaptive performance. The Dual-aware Router simultaneously considers scenario-level features and routing uncertainty to dynamically activate expert modules. Through the effective coupling of the Global Expert and the Scene-Adaptive Experts Group via the Dual-aware Router, GEMINUS achieves both adaptability and robustness across diverse scenarios. GEMINUS outperforms existing methods in the Bench2Drive closed-loop benchmark and achieves state-of-the-art performance in Driving Score and Success Rate, even with only monocular vision input. The code is available at https://github.com/newbrains1/GEMINUS.