Search

Browse

• Communities
• Authors
• Year
• Titles
• Subjects

My Space

• Login

Ship-UAVs cooperation for the restoration of marine areas

Makatsoris Konstantinos

Full record

---

Summary

The deposition in the seas of substances and materials that cause pollution of marine waters is increasing. The decomposition time of such polluting factors can exceed 500 years. This situation poses enormous dangers, affecting as they do the entire ecosystem. The present dissertation puts forward the development of a multiagent system for the exploration of marine waters in order to detect polluting substances and materials. Our main goal was to adapt existing object detection algorithms to the needs of our work, and to combine the above detection algorithms within a multiagent system consisting of autonomous agents that aim to detect all polluting substances and materials present in water. In addition, we propose the implementation of specific algorithmic solutions that enhance the operation of the multiagent system, while we have developed a fairly realistic environment to simulate the operation of our multi-agent system. In more detail, our study assumes the existence of a moving sea vessel (e.g., a ship) in areas where we wish to investigate the existence of polluting substances and materials. Autonomous Unmanned Aerial Vehicles (UAVs) depart from the ship and will be called upon to cooperate and locate the polluting factors, in order to eventually follow a process of collection and removal of pollutants. To detect the polluting factors, we used the well-known Faster R-CNN and Yolo deep convolutional neural network algorithms, adapting them to the needs of our research, and combining their results with an appropriate heuristic algorithm we created. The control as well as the monitoring of our proposed multiagent system is performed via a simulation system that we designed, which gives the UAVs freedom of movement on the x, y, z axes, and simulates the function of detection sensors to detect the state of the agents (which comprises of their position, speed, and direction), as well as the operation of the communication channels both among the agents and the communication between the agents and the ship. In addition, we developed a novel heuristic algorithm as well as one variant of the well-known Kalman Filter algorithm, for determining the state of the agents based on data obtained from sensors; while we also created a problem-oriented communication language used by the agents to communicate with each other and with the sea vessel. In addition, we utilized the well-known k-means algorithm to divide the detection area into subgroups according to the number of agents involved, and developed a heuristic algorithm to cover the area assigned to pollutant detection agents. Finally, we developed an algorithm for estimating the agents' energy consumption. The task of locating pollutants in the present work obviously takes place entirely in an aquatic environment. Nevertheless, the specific research and methodology proposed and followed, could be used with minimal changes in other settings, such as in search and rescue missions.

Available Files

Services