Using reinforcement learning to fight forest fires: Comparing CMC with CoSyNE

Roothaert, H.M. (2020) Using reinforcement learning to fight forest fires: Comparing CMC with CoSyNE. Bachelor's Thesis, Artificial Intelligence.

Preview	Text Bachelor_thesis_Graded.pdf Download (406kB) | Preview
	Text toestemming.pdf Restricted to Registered users only Download (97kB)

Abstract

Minimising both the economical and physiological damages caused by forest fires is a complex problem. This paper builds upon an existing Reinforcement Learning (RL) based Decision Support System (DSS), which could optimise fire management techniques once fully developed. The aim is to find out which RL-technique is best suited for forest fire control, Connectionist Monte Carlo (CMC) or Cooperative Synapse NeuroEvolution (CoSyNE). Both RL techniques were trained to place sub-goals optimally in all wind directions around the center of a simulated fire. These sub-goals are in turn completed by firefighting agents. By varying the number of agents and the path-finding algorithm used by these agents, different levels of complexity were induced to the fire management problem. Overall, CoSyNE significantly outperformed CMC. There are however considerable improvements to be made to both the implementation of the RL techniques and to the simulation before it can be used to train a reliable DSS.

Item Type:	Thesis (Bachelor's Thesis)
Supervisor name:	Wiering, M.A.
Degree programme:	Artificial Intelligence
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	05 Mar 2020 11:35
Last Modified:	05 Mar 2020 11:35
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/21637

Actions (login required)

View Item