Illegal, unreported and unregulated fishing is a worldwide problem that is causing local and global financial losses, depleting natural resources, changing our diverse ecosystem and causing undue pressure upon the fishing industry. This paper presents a Reinforcement-Learning-based approach to response generation once this type of fishing event has been detected. The Fuzzy Actor Critic Learning technique is used to train one or more pursuers to effectively catch an evader. This technique is utilized on both the pursuer and evader vessel agents in order to simulate real-world illegal and unreported fishing pursuit events. Simulations are executed along two such scenarios, namely the Automatic Identification System Gaps and the Local Fishing, Foreign Delivery ones involving both illegal and unreported fishing vessels (as evaders) and law enforcement vessels (as pursuers). Experimental results are presented and analyzed whereby the pursuers catch the evading fishing vessels within a preset capture time. To our knowledge, this is the first time any Reinforcement Learning techniques have been applied as a response to such fishing events. The proposed methodology here is generic enough that it can be easily extrapolated to other domains.

Additional Metadata
Persistent URL
Conference 2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017
Akinbulire, T. (Tolulope), Schwartz, H.M, Falcon, R. (Rafael), & Abielmona, R. (Rami). (2018). A reinforcement learning approach to tackle illegal, unreported and unregulated fishing. In 2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings (pp. 1–8). doi:10.1109/SSCI.2017.8285315