Distributed, Safe Robot Control Demonstrator

This is the public version of this page. Please login for more information.

This demonstrator will leverage Drona, TuLiP, and ComPlan (see below) to demonstrate on-the-fly synthesized control of aerial drones and (possibly) ground vehicles. Drona is a programming framework for distributed mobile robotics (DMR) developed by PI Seshia’s team. Drona combines a high-level programming languages for DMR applications combined with a verified software stack that runs on top of standard robotics platforms such as the Robot Operating System (ROS). Drona also includes a decentralized motion-planning protocol that interfaces with ComPlan, a novel multi-robot motion planning toolbox developed through a collaboration between PIs Kumar, Pappas, and Seshia. Drona and ComPlan have been successfully used to program teams of drones to achieve complex tasks specified in temporal logics in a coordinated fashion while providing strong guarantees of safety and correctness. Additionally, PI Murray’s group has developed the TuLiP toolbox for temporal logic based planning for robotics and control applications. TuLiP can synthesize finite-state controllers from linear temporal logic specifications. PI Mangharam’s group has developed the F1/10 Autonomous Racing platform which features 1/10th scale Formula-1 electric vehicles, which can race at maximum speeds of 40mph indoors. The platform is an excellent vehicle to teach the principles of autonomy - perception, planning, control and coordination - and also a core TerraSwarm effort to identifying pathways to safe and affordable autonomous systems. Each vehicle features LIDAR, stereo-cameras, depth sensors, IMUs and computation on a hybrid GPU/CPU platform. To scale the effort, we have developed the http://f1tenth.org website which features detailed instructions and video lectures on how to build, drive and race F1/10 autonomous vehicles. The course is currently being taught in over a dozen institutions and the bi-annual competition has attracted several hundred researchers. We will demonstrate collaborative driving, robust strategies for aggressive overtaking, driving at the limits of control with vehicles on a racetrack and an autonomous racing simulation platform for algorithmic development. Part of this demonstration will be an integration, through the SwarmOS, of third party frameworks, particularly ROS (the Robot Operating System originally developed by Willow Garage) and the ModP Language (an event-driven programming framework developed by PI Seshia’s group in collaboration with Microsoft Research).