This paper presents the basic design and evaluation of a prediction-based controller for real-time object tracking. After a system overview, the trajectory identification and prediction module is studied, presenting a prediction system for non-linear trajectories. The prediction system is based on a set of parallel linear predictors, each of which represents a hypothesis about the target trajectory. The performance of the system is compared to a traditional linear prediction model, showing an improvement of one order of magnitude in the prediction error.
展开▼
