Fault tolerant multi-sensor fusion based on the information gain

13th European Workshop on Advanced Control and Diagnosis

Al Hage, Joelle^1 ; El Najjar, Maan E.^1 ; Pomorski, Denis^1

University of Lille, Laboratory CRIStAL, UMR CNRS 9189, Villeneuve d'Ascq, France^1

关键词: Environmental Monitoring;    Exteroceptive sensor;    Information gain;    Multi-robot systems;    Multi-sensor fusion;    Natural disasters;    Sensor fault diagnosis;    Threshold optimization;   

Others  :  https://iopscience.iop.org/article/10.1088/1742-6596/783/1/012011/pdf
DOI  :  10.1088/1742-6596/783/1/012011

来源: IOP

【 摘 要 】

In the last decade, multi-robot systems are used in several applications like for example, the army, the intervention areas presenting danger to human life, the management of natural disasters, the environmental monitoring, exploration and agriculture. The integrity of localization of the robots must be ensured in order to achieve their mission in the best conditions. Robots are equipped with proprioceptive (encoders, gyroscope) and exteroceptive sensors (Kinect). However, these sensors could be affected by various faults types that can be assimilated to erroneous measurements, bias, outliers, drifts,... In absence of a sensor fault diagnosis step, the integrity and the continuity of the localization are affected. In this work, we present a muti-sensors fusion approach with Fault Detection and Exclusion (FDE) based on the information theory. In this context, we are interested by the information gain given by an observation which may be relevant when dealing with the fault tolerance aspect. Moreover, threshold optimization based on the quantity of information given by a decision on the true hypothesis is highlighted.