Bayesian-networks-based motion estimation for a highly-safe intelligent vehicle

Motion estimation of a moving object is one of the most important technologies to develop a next-generation highly-safe intelligent vehicle. Although intention of a driver in a target vehicle is key information for the motion estimation, we can not observe directly from sensors. This article presents a building method of Bayesian Networks (BNs) for motion estimation related to a driver's intention. Driver's intentions are hierarchically defined, so that the BN becomes as simple as possible. Causal relation between the intentions is discussed to reflect the real-world motion process. As a result, not only the quality of motion estimation but also the inference performance can be increased, Experimental learning system based on two-dimensional image processing is also presented for automatic acquisition of the BN probabilistic parameters.