Interaction-Merged Motion Planning: Effectively Leveraging Diverse Motion Datasets for Robust Planning

2025-07-07

AI Summary

To address domain imbalance, catastrophic forgetting, and high computational overhead arising from heterogeneous interaction patterns and environmental distributions in multi-source trajectory data for autonomous driving motion planning, this paper proposes a two-stage parameter checkpoint fusion method. First, behavioral pre-merging aggregates behavior-level knowledge across source-domain models; second, weighted parameter fusion performs model-level integration—both without fine-tuning or retraining, thereby circumventing inherent limitations of conventional domain adaptation and ensemble learning. Evaluated on multiple benchmarks (nuPlan, Argoverse 2) and state-of-the-art planners—including diffusion-based and Transformer-based architectures—the method significantly improves cross-domain generalization and planning robustness: average collision rate decreases by 23.6%, and planning success rate increases by 18.4%.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Multiagent Systems: Multiagent Planning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

Motion planning is a crucial component of autonomous robot driving. While various trajectory datasets exist, effectively utilizing them for a target domain remains challenging due to differences in agent interactions and environmental characteristics. Conventional approaches, such as domain adaptation or ensemble learning, leverage multiple source datasets but suffer from domain imbalance, catastrophic forgetting, and high computational costs. To address these challenges, we propose Interaction-Merged Motion Planning (IMMP), a novel approach that leverages parameter checkpoints trained on different domains during adaptation to the target domain. IMMP follows a two-step process: pre-merging to capture agent behaviors and interactions, sufficiently extracting diverse information from the source domain, followed by merging to construct an adaptable model that efficiently transfers diverse interactions to the target domain. Our method is evaluated on various planning benchmarks and models, demonstrating superior performance compared to conventional approaches.

Problem

• Effectively utilizing diverse trajectory datasets for robot motion planning
• Addressing domain imbalance and catastrophic forgetting in motion planning
• Transferring diverse interactions to target domains efficiently and robustly

Innovation

• Uses parameter checkpoints from diverse domains
• Two-step process: pre-merging and merging
• Efficiently transfers diverse interactions to target