Project description
As autonomous systems become increasingly capable, there is growing interest in teams of robots that collaborate to perform multiple tasks. Examples include environmental monitoring, search and rescue, warehouse automation, surveillance and autonomous vehicle coordination. A challenge in these systems is determining which robots should perform which tasks. This project will investigate learning-based approaches to multi-agent task allocation, with a particular focus on a recent multi-armed bandit decision-making algorithm. The research will explore how agents can balance exploration and exploitation to efficiently learn task assignments while adapting to changing conditions. Possible directions, depending on student interests, include:- Algorithm development and analysis.
- Investigation of performance guarantees, convergence properties and scalability.
- Implementation of a robotics simulation environment for algorithm evaluation.
- Tailoring of the simulations to a particular application of interest.
Assumed knowledge
The project is suitable for students with general interests in any of robotics, machine learning, optimisation, control systems or autonomous systems. An amount of coding experience/interest in MATLAB or Python is preferable.
Note: You need to register interest in projects from different supervisors (not a number of projects with the one supervisor).
You must also contact each supervisor directly to discuss both the project details and your suitability to undertake the project.