Croatian Science Foundation installation research project
2021 - 2026
The utilization of Unmanned Aerial Vehicles (UAVs) in search missions has many advantages, including maneuverability, human risk reduction and cost effectiveness. The application of search algorithms provides aspace to develop and use state-of-the-art methods which can greatly increase the performance of the search. When searching for a non-stationary target, such as a person floating in the sea, the consideration of its movement dynamics is critical to search success. We propose to test the applicability of a novel area coverage method and target detection and sensing system on the problem of searching in oceanic environments. For the area coverage method we intend to use HEDAC (Heat Equation Driven Area Coverage) method which has shown to be one of the most advanced algorithms for heterogeneous multi-agent control in steady conditions. HEDAC should allow the search for unsteady target distribution with the use of multiple agents. We plan to use multiple UAVs with heterogeneous sensing equipment and flight parameters. These will be rotary-wing UAVs equipped with visual and infrared spectrum cameras. Visual detection and robust tracking of targets can be accomplished by making use of supervised or semi-unsupervised deep convolutional and recurrent neural networks. The proposed research comprises two types of experiments. The first experiment would be a ground search scenario with stationary targets. In a second experiment, which is a feasible search at sea scenario, UAVs will need to detect floating targets autonomously, while following the real-time updated goal area coverage. Sea surface velocity will be continuously measured by use of scattered floating GPS-enabled buoys, which will be used for feeding a machine learning sea flow predictor system. Since conventional search relies on the presumptions of steady and uniform distribution of targets, the proposed procedure should represent a significant improvement in the search methodology.
PI: Stefan Ivić