Back to the program.
Let's Talk About Stigmergy
Carlo Pinciroli
Worcester Polytechnic Institute
On 2022-04-12 at 16:00:00 (Brussels Time)

Abstract

Stigmergy is an environment-mediated form of communication that enables large-scale coordination in collective systems. Social insects, for example, lay pheromones to enrich the environment with markers used by kin individuals to find food, build nests, and defend from predators. In this talk, I will discuss several facets of stigmergy as a coordination mechanism for robot swarms. I will first present how to implement stigmergy as a distributed data structure shared by minimalistic robots. I will present two applications of this concept: decentralized federated learning and constructing semantic maps with fallible individuals. I will then show that the concept of stigmergy can be extended to make collective construction more effective. Lastly, I will discuss how fragile stigmergic mechanisms are to malicious attacks, along with ideas to limit the damage such attacks can cause.

Short Bio of the Speaker

Carlo Pinciroli is an assistant professor at Worcester Polytechnic Institute, where he leads NEST Lab. A former IRIDIA graduate under the supervision of Marco Dorigo and Mauro Birattari, his research interests include multi-robot systems, swarm intelligence, and software engineering. Carlo published more than 50 peer-reviewed articles and 2 book chapters, and he edited 2 books. He is associate editor for the Swarm Intelligence journal and for the IROS and ICRA robotics conferences. Carlo was the recipient of several awards for research and teaching, including a best paper award at AAMAS (2022), WPI’s Romeo L. Moruzzi Young Faculty Award for Innovation in Undergraduate Education (2020) and an Amazon Research Award (2019). But for most people he's just "the ARGoS guy."

Link to Seminar:

https://teams.microsoft.com/l/meetup-join/19%3aYDVRTvS5v-FcnsTJisONwlDTQrEw-hclXJrmwzVxbkg1%40thread.tacv2/1649238457142?context=%7b%22Tid%22%3a%2230a5145e-75bd-4212-bb02-8ff9c0ea4ae9%22%2c%22Oid%22%3a%22dcfe88e1-1c78-4028-8f17-6357145735b2%22%7d