Dynamic Object Detection and Representation for Mobile Robot Application

This paper presents how to classify the sensorial information to set the label of the cells in the occupation grid map of an unknown environment where several robots are wandering and how to use these labels to complete a path execution task. Reactive navigation of a mobile robot needs to identify the objects that encounters in its way towards a specified goal. The robot does not know the environment but it knows the position of its goal. The robot needs to build a map during its navigation to the goal. A pattern classification task is needed to update the state of the map grid cells as more information is acquired by the sensors. We propose to use a finite state machine approach to set the correct label according to the objects in the environment. The occupancy grid map is crucial to complete the navigation task safely. We have tested our approach in several scenarios with a varying number of mobile obstacles. The results show that our method works efficiently in a set of benchmark scenarios of varying degree of complexity.