Accurate perception of environmental ferrous features' positioning using a simple mobile robot platform

Probabilistic perceptions are the most proper tools for building mobile robotics mapping especially if no accurate model of the sensory system and the environmental features are in hand. Here, we briefly introduce architecture of our Demining robot (DR) and then deeply investigate the scanning and mapping method it employs in order to gather the knowledge from the working environment, communicate to the central point, polish the gathered data with a proper mathematical model, and indeed generate the output such that represents the hidden ferrous features of the work space. Since electromagnetic sensors are not accurate enough to include ranging information, a probabilistic model enriched with a filtering technique has significantly enhanced the location accuracy of the mapped features to reach the quality of 99.247% in average. Full autonomy of the robot for scanning the unknown features, eases the raw-data acquisition procedure for an on/offline refinement of white thermal random electromagnetic noise on the sensors. Depending on the application, false alarm and false negative thresholds are flexible to be defined with respect to the achievable mapping resolution.