A Computationally Efficient Obstacle Avoidance System for UAV Teleoperation
Shuhan He, Gang Chen · 2021 27th International Conference on Mechatronics and Machine Vision in Practice (M2VIP) · 2021
In spite of the development of autonomy in unmanned aerial vehicles (UAV), some complex missions, which typically require cognition or intuitive decision making, are still beyond the capability of fully autonomous UAVs. Human-robot interaction (HRI) are necessary in these missions. Typical examples could be inspecting machines in factory or taking an aerial shot in cluttered environment. Furthermore, the operators in the aforementioned missions are usually inexperienced in manipulating UAVs. Existing teleoperation systems are usually designed for experienced operators and are unable to preserve purposes of the users. Therefore, it is fundamental to develop a teleoperation system which could integrate human intentions and ensure safety autonomously. In this work, a computationally efficient obstacle avoidance teleoperation system is developed and tested. A sampling-based trajectory planning method is utilized to achieve real-time planning and robust performance. Human intentions are given by a remote control. Besides, different control strategies are chosen in different environments based on distance map entropy to facilitate convenient operation. The whole system is proved to work efficiently and robustly by indoor and outdoor experiments.