Multi-Agent Task allocation for Heterogenous Salad Harvesting Teams

Multi-agent task allocation (MATA) methods distribute a set of tasks to a set of agents with the aim of minimising (or maximising) one or more objectives. In real-world settings, such as salad farms, agents are heterogeneous, and performance can be measured using different objectives/metric. Our aim is to model the uncertainty in performance of different salad harvesting agents (e.g. humans with knifes/scissors and rotary harvesting devices) using a range of different metrics (e.g. harvesting speed, amount of waste produced, energy consumed). The produced model will then be used to efficiently allocate harvesting tasks to the different agents.

The project will involve three core technology-based stages: (1) defining a set of metrics to measure the performance of different salad harvesting agents and gathering data on those metrics; (2) integrating uncertainty in agent performance into a MATA approach and evaluating the approach on the data gathered; and (3) investigating the steps required to deploy the MATA approached developed on a commercial farm - this will include performing experiments within simulation and on our Riseholme campus farm.

The PhD candidate will have the opportunity to learn about gathering and processing real-world data; and designing and conducting experiments within simulation and the physical world. They will expand the knowledge of multi-agent task allocation approaches and simulations. They will also develop their oral communication skills thorough giving talks to the research group and at conferences/workshops and their written communication skills by producing publications.