Fault recovery strategies for resilient robot swarms


Foot-bot robot swarm. Reprinted with permission by Marco Dorigo and Nithin Mathews, IRIDIA, ULB.

Robot swarms are large-scale multirobot systems with decentralized control which means that each robot acts based only on local perception and on local coordination with neighboring robots. The decentralized approach to control confers a number of benefits in harsh environments. Robot swarms have the potential to take on numerous real-world tasks. In particular, tasks that require sensing or action over large areas or at a high spatiotemporal resolution, such as environmental monitoring, are candidates for application of future swarm robotics systems.  However, robot swarms to date are frail systems, unprepared for long-term autonomy. They require hours of learning to adapt their behavior, to recover from faults inevitably sustained during operation, and are easily challenged by inevitable changes in their environments.

Swarm of builder robots. Credit: Eliza Grinnell, Harvard School of Engineering and Applied Sciences


Aquatic surface drone at BioMachines Lab Lisbon, Portugal.

In this new project we aim to address this situation by providing algorithmic approaches for low-cost robot swarms to adapt their behavior — in no more than a few minutes — to faults sustained by individual robots of the collective. The developed algorithms will be demonstrated on, ground mobile robots and aquatic surface drone swarms. Success in this project would demonstrate that swarm behavior adaptation can enable large-scale robot swarms to recover from faults, sparking further research into integrating such robotic systems into our day-to-day lives, with robot swarms working seamlessly around us, performing vital tasks such as the autonomous monitoring of large water-bodies for pollutants, using a swarm of aquatic surface drones. 

About us

RHex robot platform of the swarm

We are an internationally renowned team of researchers based at the School of Electronics and computer Science at the University of Southampton. Our research interests are focused on developing and applying techniques from machine learning, evolutionary computation, swarm robotics, multi-agent systems, and autonomous systems to a range of complex real-world problems. We have an impressive track record of  published research in top conferences and journals in these areas.


Legged robot recovering from damage

Are you excited about pushing robot swarms out of their tightly controlled laboratory environments, to operate with minimal human intervention in the real-world; to fulfill their potential to tackle high-impact problems in distributed automation, from precision agriculture on land to monitoring for pollutants in our oceans? Are you passionate about developing a novel family of machine learning algorithms for the next-generation of robot swarms — capable of adapting to sustained damages in minutes instead of hours?

We currently have a position open for a PhD researcher — if you like to join our effort, please email d.s.tarapore@soton.ac.uk with your CV and motivation letter.

Related Publications

Cully, Antoine, Jeff Clune, Danesh Tarapore, and Jean-Baptiste Mouret. “Robots that can adapt like animals.” Nature 521, no. 7553 (2015): 503.

Tarapore, Danesh, Anders Lyhne Christensen, and Jon Timmis. “Generic, scalable and decentralized fault detection for robot swarms.” PloS one 12, no. 8 (2017): e0182058.

Mathews, Nithin, Anders Lyhne Christensen, Rehan O’Grady, Francesco Mondada, and Marco Dorigo. “Mergeable nervous systems for robots.” Nature communications 8, no. 1 (2017): 439.

Bredeche, Nicolas, Evert Haasdijk, and Abraham Prieto. “Embodied Evolution in Collective Robotics: A Review.” Frontiers in Robotics and AI 5 (2018): 12.

Werfel, Justin, Kirstin Petersen, and Radhika Nagpal. “Designing collective behavior in a termite-inspired robot construction team.” Science 343, no. 6172 (2014): 754-758.

Silva, Fernando, Luís Correia, and Anders Lyhne Christensen. “Evolutionary online behaviour learning and adaptation in real robots.” Royal Society open science 4, no. 7 (2017): 160938.