Detection of moving fish schools using reinforcement learning technique

In this study, it is aimed to contribute to the fishing sector by determining the locations of moving fish schools. With the Q-Learning algorithm, areas where fish schools are frequently seen were marked and autonomous ships were able to reach these areas faster. With the Q-Learning algorithm, one of the machine learning techniques, areas where fish schools are abundant were determined and reward and penalty points were given to each region. In addition, the fish density matrix of the region was extracted thanks to the autonomous systems. Moreover, the algorithm can be automatically updated according to fish species and fishing bans. A different Q-Gain matrix was kept for each fish species to be caught, allowing autonomous ships to move according to the gain matrix. In short, high gains were achieved in terms of time and travel costs in finding or following fish schools by recognizing the region by autonomous ships.