Automatic Collision Avoidance for Manually Tele-operated Unmanned Aerial Vehicles (pdf)

Jason Israelsen, Matt Beall, Daman Bareiss, Dan Stuart, Eric Keeney, Jur van den Berg


In this paper we present an approach that aids the human operator of unmanned aerial vehicles by automatically performing collision avoidance with obstacles in the environment so that the operator can focus on the global direction of motion of the vehicle. As opposed to systems that override operator control as a last resort in order to avoid collisions (such as those found in modern automobiles), our approach is designed such that the operator can rely on the automatic collision avoidance, enabling intuitive and safe operator control of vehicles that may otherwise be difficult to control. Our approach continually extrapolates the future flight path of the vehicle given the current operator control input. If an imminent collision is predicted our algorithm will override the operator's control input with the nearest control input that will actually let the vehicle avoid collisions with obstacles. This ensures safe flight while simultaneously maintaining the intent of the human operator as closely as possible. We successfully implemented our approach on a physical quadrotor system in a laboratory environment. In all experiments the human operator failed to crash the vehicle into floors, walls, ceilings, or obstacles, even when deliberately attempting to do so.




Featured in IEEE Spectrum.