SBIR-STTR Award

Use of data from multiple sensors provides the opportunity for improved ballistic missile defense (BMD) search and tracking. Algorithms for combining multi-sensor data are required. BMD sensor data fusion is a challenging problem because of incompatibility in coordinate systems for different sensors, which makes it difficult to transfer variance and covariance information, and because of sensor registration issues, which result in measurement errors with a consistent bias. To overcome these difficulties, we propose using a root-mean-square (RMS) approach. The RMS value of a possible missile trajectory expresses the amount of consistency between the measurements that the possible trajectory should generate, and the actual measurements. The Marquardt-Levenberg optimization method can be used to find the trajectory with minimum RMS, which can be used as the fire control solution for that missile. Sensor bias can be resolved, allowing registration errors to be corrected. Furthermore, an output of the Marquardt-Levenberg algorithm is a covariance structure on the parameter set that defines possible missile trajectories.

Keywords:
Ballistic Missile Defense, Sensor Fusion, Data Fusion, Root Mean Square, Marquardt-Levenberg, Sensor Registration

Use of data from multiple sensors provides the opportunity for improved ballistic missile defense search and tracking. Algorithms for combining multi-sensor data are required. Ballistic missile defense sensor data fusion is a challenging problem because of incompatibility in coordinate systems for different sensors, which makes it difficult to transfer variance and covariance information, and because of sensor registration issues, which result in measurement errors with a consistent bias. To overcome these difficulties, we propose to use a root-mean-square approach to identify bias errors and outliers. The root-mean-square value of a possible missile trajectory expresses the amount of consistency between the measurements that the trajectory should generate, and the actual measurements. Optimization methods including the Marquardt-Levenberg algorithm will be used to find the trajectory with minimum root-mean-square. Sensor bias can be resolved, allowing registration errors to be corrected. A computer program will be developed to test sensor data streams for autocorrelation, bias errors and outliers and to remove them from the data stream. Field data and synthetic data will be used to measure the effectiveness of the sensor data fusion algorithms.

Keywords:
Ballistic Missile Defense, Sensor Fusion, Data Fusion, Root Mean Square, Bias Error, Missile Tracking, Autocorrelation Error