Multi-scale adaptive sampling with mobile robots for mapping of forest fires

The use of robotics in distributed field monitoring applications requires wireless sensors that are deployed efficiently. A very important aspect of mobile sensor deployment includes sampling algorithms at locations most likely to yield useful information about a spatio-temporal field variable of interest. This paper proposes to use robotic nodes to estimate the time-varying spread of wildfires using a distributed multi-scale adaptive sampling strategy. Our proposed algorithm, "EKF-NN-GAS", is based on neural networks, the Extended Kalman Filter and greedy search heuristics. This sampling strategy combines measurements arriving at different times and scale lengths from sensors that could be located on ground, air-borne and space-borne observation platforms. We present the mathematical formulation of the algorithm directing single and multiple robots to reconstruct a spatio-temporal forest fire spread. Simulation results show that adding search and classification heuristics to the sampling strategy significantly improves the field reconstruction time, and can lead to an efficient implementation with multiple fire-tracking robots.