[论文解读] Physical Interaction and Manipulation of the Environment using Aerial Robots

Azarakhsh Keipour|arXiv (Cornell University)|Jul 6, 2022

Robotic Path Planning Algorithms被引用 6

一句话总结

本文提出了一种用于全驱动多旋翼飞行器的实时控制器，可实现与环境的精确物理交互，包括可变形物体。该方法提出了一种估算机器人瞬时可用全部作用力与力矩的方法，提升了控制性能，并使此前被认为不可行的操控任务成为可能。

ABSTRACT

The physical interaction of aerial robots with their environment has countless potential applications and is an emerging area with many open challenges. Fully-actuated multirotors have been introduced to tackle some of these challenges. They provide complete control over position and orientation and eliminate the need for attaching a multi-DoF manipulation arm to the robot. However, there are many open problems before they can be used in real-world applications. Researchers have introduced some methods for physical interaction in limited settings. Their experiments primarily use prototype-level software without an efficient path to integration with real-world applications. We describe a new cost-effective solution for integrating these robots with the existing software and hardware flight systems for real-world applications and expand it to physical interaction applications. On the other hand, the existing control approaches for fully-actuated robots assume conservative limits for the thrusts and moments available to the robot. Using conservative assumptions for these already-inefficient robots makes their interactions even less optimal and may even result in many feasible physical interaction applications becoming infeasible. This work proposes a real-time method for estimating the complete set of instantaneously available forces and moments that robots can use to optimize their physical interaction performance. Finally, many real-world applications where aerial robots can improve the existing manual solutions deal with deformable objects. However, the perception and planning for their manipulation is still challenging. This research explores how aerial physical interaction can be extended to deformable objects. It provides a detection method suitable for manipulating deformable one-dimensional objects and introduces a new perspective on planning the manipulation of these objects.

研究动机与目标

• 开发一种成本低廉的控制器，可与现有飞行系统集成，用于现实世界中的空中操作。
• 解决全驱动多旋翼飞行器中保守推力与力矩假设导致的效率低下问题。
• 实现与一维可变形物体（如电缆或绳索）的物理交互。
• 研究在三维空间中操控可变形物体的可行性及规划需求。
• 弥合原型研究与可部署现实应用之间的差距。

提出的方法

• 提出一种实时算法，用于估算全驱动多旋翼飞行器瞬时可用的全部作用力与力矩。
• 提出一种新型输入映射矩阵，将控制输入与机器人动力学解耦，实现控制器设计。
• 基于旋翼几何结构与角速度，利用多旋翼的几何模型计算单个旋翼产生的推力与力矩贡献。
• 通过将系统模型扩展以包含外部 wrench（广义力），使控制器能够处理末端执行器的交互。
• 开发一种感知方法，利用机载传感器检测并跟踪一维可变形物体。
• 提出一种新的规划视角，综合考虑物体动力学、末端执行器精度及三维空间中所需的 wrench。

实验结果

研究问题

• RQ1 全驱动多旋翼飞行器如何通过实时准确估算可用作用力与力矩，实现最优物理交互性能？
• RQ2 何种控制架构可实现与现有飞行软件和硬件系统的无缝集成，以支持现实世界部署？
• RQ3 空中机器人如何有效操控一维可变形物体（如电缆或绳索）？
• RQ4 在三维空间中操控可变形物体时，关键的动力学与运动学约束是什么？
• RQ5 为实现空中机器人对可变形物体的可靠操控，需要哪些感知与规划策略？

主要发现

• 所提出的实时力与力矩估算方法，使物理交互比保守假设下更加高效且可行。
• 控制器设计成功与现有飞行系统集成，实现了全驱动多旋翼飞行器在现实任务中执行操控的部署。
• 检测与跟踪一维可变形物体的方法在动态环境中表现出可靠的性能。
• 结合动态物体模型与精确的 wrench 估算进行规划，相比刚性物体假设，显著提升了操控成功率。
• 该系统在模拟与真实场景中均验证了在电缆操控及与可变形结构交互方面的可行性。
• 本研究为未来在真实环境中实现对复杂柔性材料的自主空中操控奠定了基础。