Ecologically-Inspired Multi-Robot Systems



Recent advances in robotics and control systems have motivated a push for long-duration autonomy, where agents are left to operate in uncertain environments for days, weeks, or even months at a time. At the same time, the miniaturization of computer hardware has made the deployment of multi-agent systems for engineering applications economically feasible. These factors have motivated the development of a new framework for the design of autonomous systems: ecologically-inspired robotics. In this tutorial, we will cover the theoretical foundations, software implementation, and several recent case studies that apply the ecologically-inspired robotics framework to control problems. Under this framework, agent interactions, tasks, and safety requirements are embedded as constraints in an optimization problem where each agent minimizes its energy consumption. The resulting agent behavior is interpretable and data-driven; agent actions are determined by the active constraints, and the boundary of the feasible action space is constructed through sensing and communication. Problems in this domain are closely related to control barrier functions and viability theory; many are amenable to real-time algorithms, including quadratic programming. This enables many ecologically-inspired robotics problems to be solved efficiently using modest hardware, and as an added benefit, the resulting system behavior is robust to the addition, removal, and failure of agents during operation. The tutorial will consist of two parts: a series of lectures, and two interactive activities. The organizers will present an overview and motivation for ecologically-inspired robotics, which will be followed by a theoretical overview of the topic and several case studies with implementation details. We will close the tutorial with a coding challenge and experimental demonstration to foster discussion and give attendees the opportunity to apply the framework directly.

Coding Tutorial

The coding tutorial uses Python3 with the QPsolvers package to implement a multi-agent go-to-goal problem in Google Colab

We have applied ecologically-inspired control to a variety of applications with code available online:

Tentative Program


Assistant Professor

Old Dominion University

Assistant Professor

University of Waterloo