Arm2Air: Cross-Embodiment Skeleton Transfer for 3D Relay Formation

2026-07-29

AI Summary

This work addresses the challenge of deploying drone relay networks in complex urban 3D environments, where line-of-sight blockages, communication range limitations, flight altitude constraints, and obstacles severely hinder performance. To overcome these issues, the authors propose a novel cross-heterogeneous embodied task transfer paradigm: leveraging the structured action sequences from a pretrained robotic arm obstacle-avoidance policy as prior knowledge, they employ a Transformer-based transfer framework with LoRA for efficient adaptation. Using only three target environment maps, the method initializes and optimizes relay links with remarkable data and computational efficiency. In highly occluded urban scenarios, it achieves a 32.6% improvement in bottleneck capacity, reduces planning time by 64.9%, decreases relay positioning error by 53.6%, and requires updating merely 0.134 million parameters.

Abstract

Unmanned aerial vehicle (UAV) relay networks can restore connectivity after communication infrastructure is damaged. Urban relay placement is difficult because line-of-sight blockage, communication range, altitude, and three-dimensional obstacles must be considered jointly. Arm2Air transfers obstacle-avoidance skeletons from robot arms to UAV relay placement through cross-embodiment transfer. Source-domain robot-arm motions from a pretrained Neural MP model are converted into ordered skeletons that pretrain a transformer-based transfer platform, which is then adapted to the UAV domain using limited target data and Low-Rank Adaptation. The transferred skeleton initializes a relay chain that is refined for connectivity, bottleneck capacity, delay, and movement cost. On nine held-out high-clutter 3D urban maps, Arm2Air reduced median end-to-end planning runtime by 64.9 percent relative to the fastest conventional planner. On the high-obstruction group of a separate 30-map dense urban holdout, it increased bottleneck capacity by 32.6 percent, reduced capacity variance by 74.7 percent, reduced maximum hop distance by 13.2 percent, reduced hop-distance variance by 75.2 percent, and reduced relay displacement by 16.9 percent relative to IMPC-MD. With only three target-domain training maps, Arm2Air reduced relay-position root mean square error by 53.6 percent relative to training from scratch while updating 0.134 million parameters, compared with 1.383 million for Scratch and Full Fine-tuning. These results demonstrate computationally and data-efficient UAV relay placement and suggest a broader principle for transferring ordered structural priors across heterogeneous embodied tasks.

Problem

• UAV relay placement
• 3D urban environment
• obstacle avoidance
• cross-embodiment transfer
• connectivity restoration

Innovation

• cross-embodiment transfer
• skeleton-based planning
• low-rank adaptation
• UAV relay placement
• structured prior transfer