This is my thesis from the mphil year, where I worked with researchers from the Prorok Lab at the University of Cambridge on this work. In the thesis, we presented a utilisation-aware Multi-Agent Path Planning(MAPP) environment optimisation framework. It acted as a probe to the MAPP congestion problem from utilisation perspective and thoroughly explored the trade-off between roadmap utilisation, space efficiency and flowtime. Here, we experimented with:
- Using a directed Voronoi diagram instead of Grid as the base roadmap
- Utilising Gaussian Process with Bayesian Optimisation to optimise the direction and usage preferences of edges in the roadmap
Please see Example.ipynb. Refer to Elim_MPhil_thesis.pdf for more details.