The emergence of indoor aerial robots holds significant potential for enhancing construction site workers' productivity by autonomously performing inspection and mapping tasks. The key challenge to this application is ensuring navigation safety with human workers. While navigation in static environments has been extensively studied, navigating dynamic environments remains open due to challenges in perception and planning. Payload limitations of unmanned aerial vehicles limit them to using cameras with limited fields of view, resulting in unreliable perception and tracking during collision avoidance. Moreover, the unpredictable nature of the dynamic environments can quickly make the generated optimal trajectory outdated. To address these challenges, this paper presents a comprehensive navigation framework that incorporates both perception and planning, introducing the concept of dynamic obstacle intent prediction. Our perception module detects and tracks dynamic obstacles efficiently and handles tracking loss and occlusion during collision avoidance. The proposed intent prediction module employs a Markov Decision Process (MDP) to forecast potential actions of dynamic obstacles with the possible future trajectories. Finally, a novel intent-based planning algorithm, leveraging model predictive control (MPC), is applied to generate safe navigation trajectories. Simulation and physical experiments demonstrate that our method enables safe navigation in dynamic environments and achieves the fewest collisions compared to benchmarks.

翻译：室内飞行机器人的出现，通过自主执行巡检与测绘任务，在提升建筑工地工人生产力方面展现出巨大潜力。该应用的核心挑战在于确保与人类工作者共处时的导航安全性。尽管静态环境中的导航已得到广泛研究，但由于感知与规划方面的挑战，动态环境中的导航问题仍未得到解决。无人机的有效载荷限制使其只能搭载视野有限的相机，导致避障过程中的感知与跟踪不可靠。此外，动态环境的不可预测性会迅速使生成的最优轨迹失效。为应对这些挑战，本文提出了一种融合感知与规划的完整导航框架，引入了动态障碍物意图预测的概念。我们的感知模块能高效检测并跟踪动态障碍物，并处理避障过程中的跟踪丢失与遮挡问题。所提出的意图预测模块采用马尔可夫决策过程（MDP）来预测动态障碍物的潜在动作及其可能的未来轨迹。最后，应用一种基于模型预测控制（MPC）的新型意图规划算法来生成安全导航轨迹。仿真与物理实验表明，相较于基准方法，我们的方法能够在动态环境中实现安全导航，并达到最低的碰撞次数。