Contact Name: Prof. Giuseppe Notarstefano
About the speaker
Pietro Pierpaoli is a postdoctoral fellow at the School of Electrical and Computer Engineering at Georgia Institute of Technology. He received his M.S. in aeronautical engineering from the Politecnico di Milano in 2013, and Ph.D. in mechanical engineering from the University of Miami in 2017. Pietro was a visiting scholar at the School of Electrical and Computer Engineering at Georgia Institute of Technology in 2015, and at the William E. Boing department of Aeronautics and Astronautics at the University of Washington in 2016. Pietro’s current research interests lie in the overlap of multi-agent robotics and networked control. He is particularly interested in the implementation of networked robotic systems operating in real-world environments, focusing on safety, fault tolerance, and composition of behaviors.
Abstract
One of the paradigmatic approaches in the control of multi-agent systems is formulated in terms of local rules, where the action of a single agent is defined as the gradient descent of an appropriately designed potential that depends only on the state of those neighbors the agent itself can interact with. In this way, the desired global behavior conveniently emerges from the team-wise execution of these local rules. However, the advantages of these controllers are limited by the fact that interesting applications rarely require robots to execute one single behavior indefinitely over time. On the other side, the solution of complex missions can conveniently be achieved by defining appropriate sequences of behaviors the robots should go through. For a sequence of behaviors to be correctly executed, a number of conditions need to be satisfied, e.g., the configuration of the robots at the end of each behavior needs to satisfy the configuration needed for the following one to start. As I will describe in this talk, by intersecting graph theory and control barrier function, it is possible to satisfy these constraints in a provably correct manner, and extend the versatility of multi-agent robotics in real-world applications.