Figure 1: Marine search and rescue

“We are sinking fast. Position ten miles south of San Remo…” Mayday from yacht Winston Churchill, April 1959.
Consider a possible marine disaster scenario where a ship is sinking in a storm and the crew and passengers need to be rescued before their survival expectancy vanishes. The only actions that the crew and passengers can take are to send a last mayday from the sinking ship, get on life rafts and wait, drifting due to strong winds and high waves, until a rescue team arrives. The goal of SAR in this scenario is to search for and find life rafts and rescue victims as efficiently and safely as possible.

A practically implementable robotics approach to this scenario is the use of a team of fast autonomous Unmanned Aerial Vehicles (UAVs) together with a team of autonomous helicopters with rescuers onboard. The cooperation of autonomous vehicles including fast UAVs can dramatically improve the efficiency of the search operation. In addition, if the UAVs can circle around the found life rafts to provide high beam lights and environmental information to the rescue helicopter while rescuers are retrieving victims, the efficiency and safety of the rescue operation will also improve significantly.

Figure 2: Coordinated control of heterogeneous autonomous UAVs using Bayesian filter

This project is aimed at developing a coordinated control technique that enables heterogeneous vehicles to autonomously search for and track multiple targets using recursive Bayesian filtering. Based on a filtering method, the proposed technique stochastically updates the Probability Density Functions (PDFs) of the target from their prior distributions and stochastic motion models. In addition, the use of the recursive Bayesian Filter allows nonlinear stochastic target models to be handled in the presence of non-Gaussian noise. A unified sensor model and a unified objective function are proposed to enable both search and tracking within the recursive Bayesian framework. The further strength of the proposed technique is that a vehicle can switch its task mode between search and tracking without discarding information collected during the operation.

Figure 3: 4 manned autonomous helicopters and 4 fast UAVs are search for and rescue 6 life rafts.

Figure 3 shows the cooperation of four manned autonomous helicopters and four fast autonomous UAVs to search for and rescue six life rafts. In the figure, the last life raft is about to be found. Numerical investigations have shown that the proposed technique could improve search and rescue performance by 70% on average.

Currently, a hardware-in-the-loop simulator, which allows the proposed coordinated search-and-rescue using the real hardware components, is being developed. In order to enable exploration, non-regular grid approaches, including the element-based method, are developed and also being implemented into the hardware-in-the-loop simulator.