Innovation
Lives are lost every year from people falling overboard. A new autonomous drone aims to swiftly deliver a life jacket that can buy a person time for a rescue boat to arrive.
"Our methodology incorporates different search methods and the vessel’s real-time data related to wind and currents to accurately predict a person’s location."
Traditional rescue operations work off a system where the rescue boat scours the water in triangular patterns along the most likely route that the person would have drifted. However, the boat’s low vantage point can make it hard to spot a person in the distance.
Dimosthenis Angelis’ work on the project has involved building a drone from scratch and developing algorithms that allow the drone to quickly take off and autonomously pick the best route to search for the likely location a person might have drifted to.
The drone combines a handful of computer-based search methods that have been developed to figure out likely trajectories. They take into account the uncertainties of the maritime environment and factor in the time that has lapsed since the person fell.
“Our methodology incorporates different search methods and the vessel’s real-time data related to wind and currents to accurately predict a person’s location while ensuring the drone doesn’t waste time searching the same area twice,” Dimosthenis Angelis explains.
The outcome is a novel position estimator with a much-increased success rate.
Facts
The drone’s speed and its elevated vantage point along with its smart method of searching help to increase the chances of locating people compared with conventional rescues. Tests results so far show that the drone would be able to locate more than 80 percent of people in distress.
Dimosthenis Angelis and his colleagues have recently been working on new ways for the drone to land swiftly and autonomously on return. Traditionally to land on a moving object, such as a swaying ship, drones first need to predict how the landing area would tilt to calculate a safe landing. This means lots of calibration while hovering in the air which eats up potential search time.
He reports that work on a model that allows the drone to land solely based on what the cameras see in real time is well underway:
“Having to wait up to five minutes to land is a big hit on the battery life. With our new method that we are developing, landing takes about three seconds.”
The drone has been tested on land as well as over the Kattegat, where it searched for a humanoid mannequin wearing a heated vest. Testing is ongoing and test results are used to fine-tune the equipment as well as to verify its search accuracy.
Dimosthenis Angelis’ dream scenario is that with documented results to verify its ability to locate a man overboard, every ship that carries people will install a rescue drone for extra safety.
However, he is aware that introducing new products into the heavily regulated maritime sector requires lots of work and extensive testing. In the meantime, he envisages that the drones could be adopted by the coast guard to swiftly get life jackets to people in distress.
“If we can reach a person within minutes instead of tens of minutes, we fundamentally change their chances of survival”, says Dimosthenis Angelis.
Dimosthenis Angelis PhD Student Department of Electrical and Photonics Engineering diang@dtu.dk
Evangelos Boukas Associate Professor Department of Electrical and Photonics Engineering Mobile: 5290 1729 evanb@dtu.dk