CONIX Publication

We present an example of how perception algorithms can be combined with modern control techniques to facilitate the safe control of physical systems in the presence of humans. In this demo, we task a drone with exploring our lab space while also avoiding virtual obstacles displayed in ARENA. These obstacles represent humans located in a separate lab. The position and shape of the obstacles is dynamic and corresponds of the position and orientation of people they represent. The drone avoids the obstacles by implementing a safe control policy that leverages Robust Controlled Invariant Sets (RCIS). This controller is susceptible to perception delays. Thus, we also visualize the effect of such delays on the controller’s performance. Taken as a whole, this demo showcases the increased functionality that can be achieved, as well as some inherent challenges, by utilizing intelligent perception infrastructure for the purpose of control.