Nonlinear 3-D Motion Simulation Using Multi-Sensor Data Fusion

It is a interesting topic of tracking a 3D object and estimating its 3-D motion in a multi-sensor observation environment. The motion with constant translation velocity and constant rotation rate is assumed and tracked by two separate cameras. Hence the problem considered here also involves a track-to-track fusion system in which twosensors are tracking the same target. Due to the nonlinear nature of both plant and measurement processes for the estimation system, the recursive algorithm is used to track the object and estimate its motion. Especially, an Extended Kalman filtering (EKF) is applied to create the local track for each sensor. We described the observation method and presented several track fusion approaches under different communication conditions. In centralized structure, synchronized sensors and asynchronous sensors are discussed. For hierarchical architecture, we considered two cases: with feedback and without feedback. Two communication methods (double talk & single talk) are employed in distributed system. The covariance matrix fusion is hired in both hierarchical and distributed architecture. In hierarchical system without feedback, the information filter is also approached for comparison. Various communication rates are applied to hierarchical and distributed system. The performance of track fusion is evaluated and compared to the single sensor based tracks. Index Terms — 3-D motion estimation, track-to-track fusion, Extended Kalman filtering, multi-sensor data fusion, performance evaluation