复杂工况下飞行器舱段自动对接系统的实验研究与性能评估

Research on Automated Cabin Docking System for Aircraft Based on Machine Vision

摘要:

针对固定翼飞行器舱段自动化装配过程中存在的非结构化环境干扰、初始位姿偏差大及接触力不可控等问题，对基于多模态感知的舱段自动对接系统进行了实验研究与性能评估。设计了包含视觉引导粗对准、激光测距自适应进给及力觉反馈柔顺锁定的三阶段闭环控制策略。在视觉算法方面，引入了基于注意力机制改进的U-Net模型，增强了复杂光照条件下的特征提取鲁棒性。在控制策略方面，采用基于阻抗控制的力/位混合控制，实现了接触过程的主动柔顺。为验证系统鲁棒性，设计了涵盖光照突变、动力学震动扰动及极限初始偏差的综合压力测试。实验结果表明：系统在复杂光照环境下能够保持高精度特征提取能力；在极限偏差工况下具备完全的捕获与无超调收敛能力；柔顺控制策略有效降低了接触冲击力。经验证，所提出的多模态融合与分层控制策略能够有效克服复杂工况干扰，满足飞行器舱段高精度自动化对接的工程应用需求。

Abstract:

Addressing challenges such as unstructured environmental interference, significant initial pose deviations, and uncontrollable contact forces during the automated assembly of fixed-wing aircraft cabin sections, we conducted experimental research and performance evaluations on a multi-modal perception-based automatic docking system. We designed a three-stage closed-loop control strategy, incorporating visual-guided coarse alignment, adaptive laser-ranging feeding, and force-feedback compliant locking. For visual processing, we introduced an attention-enhanced U-Net model, improving feature extraction robustness under complex lighting. Regarding control, we implemented impedance-based force/position hybrid control to ensure active compliance during contact. To assess system robustness, we designed a comprehensive stress test that simulated sudden lighting changes, dynamic vibrations, and extreme initial deviations. Experimental results demonstrated that the system maintained high-precision feature extraction in complex lighting, achieved complete capture with zero overshoot under extreme deviations, and effectively reduced contact impact forces through compliant control. Our proposed multi-modal fusion and hierarchical control strategy has proven effective in overcoming complex operational interferences, fulfilling the engineering requirements for high-precision automated docking of aircraft cabin sections.