A new framework of distributed multisensor fusion for intelligent systems

Abstract

To tackle the complexity of robotic multiple sensor fusion a suitable system structure is urgently required. Most structures at present have a fusion center to deal with information from various sensors located at different points, but unfortunately have no necessary connections between low-level sensors. Consequently the fusion center has more responsibility for communication and signal interpretation, which causes bottle-neck problem. Moreover sensors cannot help each other to improve the task environment recognition efficiency. To overcome the above drawbacks we define the concept of smart sensor at first. Each smart sensor should have five abilities, that is prediction, planning, updating, communication and assimilation. Then we discuss the algorithm of a perception system which consists of several smart sensors. At last we take two application case study as examples to illustrate our new idea.