DTCCL: Disengagement-Triggered Contrastive Continual Learning for Autonomous Bus Planners

2025-12-21

AI Summary

Autonomous buses operating in open urban environments suffer from geographically concentrated, strategy-level failures in highly interactive regions; sparse human接管 data leads to severe overfitting in conventional imitation learning. Method: We propose a接管-driven contrastive continual learning paradigm: human接管 events trigger a cloud-edge collaborative closed loop; unsupervised, geography-aware policy representation optimization is achieved via causal-aware adversarial sample generation and route-context-preserving sampling; multi-agent scenario perturbation modeling further enhances generalization. Contribution/Results: Evaluated on real-world urban bus routes, our approach improves planning success rate by 48.6% over baseline retraining methods—demonstrating the first fully autonomous, scalable, strategy-level closed-loop evolution without manual intervention.

Technology Category

• Search and Optimization: Learning to Search
• Planning, Routing, and Scheduling: Learning for Planning and Scheduling
• Intelligent Robots: Learning & Optimization for ROB

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Autonomous buses run on fixed routes but must operate in open, dynamic urban environments. Disengagement events on these routes are often geographically concentrated and typically arise from planner failures in highly interactive regions. Such policy-level failures are difficult to correct using conventional imitation learning, which easily overfits to sparse disengagement data. To address this issue, this paper presents a Disengagement-Triggered Contrastive Continual Learning (DTCCL) framework that enables autonomous buses to improve planning policies through real-world operation. Each disengagement triggers cloud-based data augmentation that generates positive and negative samples by perturbing surrounding agents while preserving route context. Contrastive learning refines policy representations to better distinguish safe and unsafe behaviors, and continual updates are applied in a cloud-edge loop without human supervision. Experiments on urban bus routes demonstrate that DTCCL improves overall planning performance by 48.6 percent compared with direct retraining, validating its effectiveness for scalable, closed-loop policy improvement in autonomous public transport.

Problem

• Addresses planner failures in autonomous buses during disengagement events.
• Improves policy learning using contrastive continual learning from sparse data.
• Enables scalable, closed-loop policy updates without human supervision.

Innovation

• Disengagement-triggered cloud-based data augmentation for policy learning
• Contrastive learning refines representations to distinguish safe behaviors
• Cloud-edge continual updates enable unsupervised closed-loop improvement