Robotic heterogeneous multi-sensor fusion with spatial and temporal alignment

The authors address the development of a coherent framework suitable for the treatment of the heterogeneous multisensor fusion problem, in the context of a robotic environment. The analysis, is based on a geometric description of the robotic environment, and concentrates on parameterizable features of the assumed rigid dynamic objects of interest present in the environment. A quantitative representation of a sensor's inherent ability to extract pertinent information from the environment is stressed. A highly efficient, flexible, and fault-tolerant, decentralized sensor fusion architecture based on a linear information type structure is presented and compared with previous work in this area. The inclusion of techniques to cope with generalized spatial and temporal uncollocation, along with conclusive discussions on the most appropriate level in the fusion structure for these alignment procedures to be performed, represent the principal contribution.<>