Variable autonomy and shared control

Lead academic:  Professor Rustam Stolkin, Professor Manuel Giuliani
Partner lead: University of Birmingham,  University of West England, Bristol

Most existing robotic systems remain fixed at one level of autonomy (LOA), typically teleoperation or autonomous control. In contrast, our recent work on variable autonomy show that the ability to dynamically switch between different LOA significantly outperforms any single LOA, e.g. pure teleoperation or pure autonomy. Alternatively, shared control frameworks enable both human and AI to collaborate to control the robot, simultaneously fusing reasoning from both the human and the autonomous agent. Such methods are particularly relevant to nuclear operations – safety-critically demands a human-in-the-loop, but, e.g. complex manipulations may be too difficult to directly tele operate efficiently. WP9 explores smooth transitions between a variety of LOA, including teleoperation, shared control, human supervised autonomy and full autonomy, including autonomous switching of LOA.

Photo shows: State-of-the-art variable autonomy system at University of Birmingham (funded by MoD). Dynamic switching of control between AI (autonomous navigation) and human (joystick) Both human and AI can in initiate changes in level of autonomy (LOA).