Quantitative Object Motion Prediction By An Adaptive Resonance Theory (ART) Neural Network

An Adaptive Resonance Theory (ART) neural network is applied for the estimation and prediction of object motion states in real time. A bottom-up process of the network keeps track of the motion history of the object and a top-down process generates the prediction of the object motion. A retrospective enforcement process adjusts the network parameters to respond dynamically to the object motion. The process does not require any assumption of the object motion model and is applicable to a variety of situations where object motion exhibits irregular and abrupt variations.