Multi Sensor Data Fusion With Risk Assessment

In robotic field, autonomous navigation safely is a timely subject. Therefore, many researchers try to elaborate a reliable system to respond this exigency. Among these systems, we mention multi sensor data fusion. The data fusion is an active field of research, It is almost applied in all domains. it is resolved by several manner. The best represented model of multi sensor fusion by excellence is the human fusion using the five senses. However, in our paper this problem is tackled based on an interesting method applied often in the risk management strategy. In our work, we have designed a new scheme of multi sensor data fusion using risk assessment of a three suggested sensors GPS, INS and Mono-SLAM. Based on these risks, we generate a three parameters dynamically used within an adaptive complementary filter. The aim of our work is to obtain an estimated pose from the three poses provided by the three sensors, but more safely than using ordinary methods of fusion. The simulation of the proposed solution shows a promising results in severe situations. Anyway, the fusion based on this approach at least gives a similar result to the ordinary complementary filter. However, the biggest improvement of the risk assessment reflects amelioration on the pose quality.