Sequence of Expert: Boosting Imitation Planners for Autonomous Driving through Temporal Alternation

2025-12-15

AI Summary

Autonomous driving imitation learning suffers from performance degradation in closed-loop settings due to cumulative temporal errors; existing approaches primarily enhance per-timestep state robustness, neglecting explicit modeling of continuous-time error propagation. This paper proposes the Sequence of Experts (SoE), a novel temporal alternating expert strategy that— for the first time—models error propagation along the continuous-time dimension. SoE achieves temporal-level robustness via lightweight, dynamic expert rotation without increasing model capacity or requiring additional annotations. It comprises three key components: a closed-loop feedback-driven expert switching mechanism, a multi-expert coordination scheduling module, and a plug-and-play architecture. Evaluated on the nuPlan benchmark, SoE consistently improves the closed-loop performance of diverse imitation planners, achieving state-of-the-art results while enhancing training efficiency and generalization stability.

Technology Category

• Planning, Routing, and Scheduling: Temporal Planning
• Machine Learning: Mixture of Experts (MoE)
• Search and Optimization: Learning to Search

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Imitation learning (IL) has emerged as a central paradigm in autonomous driving. While IL excels in matching expert behavior in open-loop settings by minimizing per-step prediction errors, its performance degrades unexpectedly in closed-loop due to the gradual accumulation of small, often imperceptible errors over time.Over successive planning cycles, these errors compound, potentially resulting in severe failures.Current research efforts predominantly rely on increasingly sophisticated network architectures or high-fidelity training datasets to enhance the robustness of IL planners against error accumulation, focusing on the state-level robustness at a single time point. However, autonomous driving is inherently a continuous-time process, and leveraging the temporal scale to enhance robustness may provide a new perspective for addressing this issue.To this end, we propose a method termed Sequence of Experts (SoE), a temporal alternation policy that enhances closed-loop performance without increasing model size or data requirements. Our experiments on large-scale autonomous driving benchmarks nuPlan demonstrate that SoE method consistently and significantly improves the performance of all the evaluated models, and achieves state-of-the-art performance.This module may provide a key and widely applicable support for improving the training efficiency of autonomous driving models.

Problem

• Addresses imitation learning's closed-loop performance degradation in autonomous driving
• Proposes a temporal alternation policy to enhance robustness without extra resources
• Improves autonomous driving models' training efficiency and state-of-the-art performance

Innovation

• Sequence of Experts alternates planners temporally
• Enhances closed-loop performance without larger models
• Leverages temporal scale to reduce error accumulation