Search papers, labs, and topics across Lattice.
This paper addresses the challenge of simultaneously achieving accurate motion prediction and safe motion planning in social robot navigation by introducing a novel framework called Disjoint Parameter Training (DPT). The authors identify a critical issue known as Skill Conflict, where overlapping parameter assignments hinder the model's ability to specialize in distinct tasks. By employing a model-merging approach that separates key parameters for each task and selectively merges only the most influential ones, DPT significantly enhances performance on standard crowd navigation benchmarks, demonstrating its effectiveness in resource-constrained environments.
Skill Conflict can severely hinder robot navigation performance, but Disjoint Parameter Training offers a solution that enhances both prediction and planning in crowded environments.
Accurate motion prediction of surrounding agents and safe motion planning are two closely coupled key tasks for social robot navigation in crowded environments. Deploying these systems on resource-constrained edge devices necessitates compact, unified models that can perform both tasks simultaneously. However, within these compact shared encoders, recent unified models often overlook severe representational conflicts that arise from the distinct objectives of predicting neighbor behaviors versus ego-centric safety planning. To address this issue, we first identify the Skill Conflict$\unicode{x2014}$a phenomenon where overlapping parameter assignments cause distinct tasks to compete for the same weights, preventing the model from fully specializing in individual skills. To resolve this, we propose a novel model-merging-based framework, Disjoint Parameter Training (DPT). DPT mitigates performance degradation caused by Skill Conflict through distributed parameter learning, which separates the key parameter regions of each task while preserving their core capabilities prior to merging. In addition, we observe that sparse merging, which selectively integrates only the most influential parameters for each task rather than combining all task-specific parameters, yields optimal performance by preventing interference among adjacent features and concentrating representational capacity. DPT can be applied in parallel with a variety of merging methods. Evaluated on standard crowd navigation benchmarks (JRDB and JTA), our framework demonstrates superior performance, validating its versatility and effectiveness for safe, resource-efficient robot navigation.
