Improved algorithm for navigation of rescue robots in underground mines

Mine rescue robots play a vital role during rescues in underground mine disasters. In this paper, we propose a new navigation method by using diverse-sensor data fusion with an improved algorithm of the Neural Network Extended Kalman Filter. During this process, we take into account that a rescue's effectiveness is limited by its single navigation model. First, we utilize the Back Propagation neural network to improve the data matching level of dissimilar sensors. Second, data fusion is carried out by combining the Extended Kalman Filter and the Back Propagation neural network. By doing so, we simultaneously retrain the Back Propagation neural network with the modified error signals. The experimental analysis showed that the algorithm can effectively deal with heterogeneous data fusion. It can also improve the convergent speed and time response of the algorithm, and further improve the accuracy of navigation. (C) 2013 Elsevier Ltd. All rights reserved.