John E. Boyd

Image fusion for tracking manoeuvring targets

It is known that improvements in target tracking can be achieved by using multiple sensors. Most commonly, the individual measurement sequences are merged using a variant of linear algorithms. The approach proposed here differs from the conventional one in that nonlinear methods of data fusion are proposed to account for the peculiarities of the different measurement categories. This technique, called complementary fusion, is illustrated with the problem of tracking an agile target. It is shown that complementary fusion not only leads to higher fidelity tracking, but it also permits the more efficient utilization of the primary sensor.

Sensor fusion in estimation algorithms

Abstract In a multimodal, system, the growth in the number of possible modal paths makes state estimation difficult. Practical algorithms bound complexity by merging estimates that are conditioned on different modal path fragments. Commonly, the weight given to these local estimates is inversely related to the normalized magnitude of the residuals generated by each local filter. This paper presents a novel dual-sensor estimator that uses a merging formula that is based upon a different function of the residuals. Its performance is contrasted with an estimator using a single sensor and with another dual-sensor algorithm that requires fewer on-line calculations.

Multiple model methods in path following II

Path following is difficult when the observation rate is low. Multiple model estimators incorporating multisensor fusion have proven useful in this application. This paper shows the advantage of a recently developed multiple model algorithm. Performance comparisons with some current algorithms are presented.

Jump-diffusion processes in tracking/recognition

Advances in sensor technology permit more sophisticated tracking/identification algorithms to be implemented. This correspondence compares and contrasts the modeling framework employed in two recent image-enhanced trackers and generalizes one of them (the PME) for use in target identification.

Brief Control of linear jump systems in noise

The equations of state evolution of a linear-jump system have a hybrid structure in which a modal-state process indexes the model of dynamic evolution of the base state. When the state measurements are noisy, the intrinsic nonlinearities in the system make it difficult to achieve acceptable performance with control algorithms based upon conventional estimators. Effective control reconfiguration requires that the correlation between the state and modal errors be computed. This paper presents an estimation algorithm that can be used in this context. It is shown by an example that closed-loop performance is improved when certain cross moments are utilized.

Data fusion using multiple models

Multiple model fusion is useful in applications in which the model of the signal processes is not known with certainty. This paper compares two current fusion algorithms with a novel alternative. The new fusion approach is shown to give improved performance when the observation rate is slow as compared with the important time constants of the signal.

New merging formula for multiple-model trackers

The quality of multiple model estimators can be improved with multisensor fusion. This paper contrasts the performance of three multiple model algorithms. It is shown that the simplest is adequate in high signal-to-noise environments. The more sophisticated warrant attention when the observations are ambiguous.

Tracking a maneuvering target using sensors of variable quality

In an environment subject to sudden change, the accuracy of tracking and prediction is strongly influenced both by the sensor architecture and by the quality of the sensors. An image-enhanced algorithm is presented for both path following and covariance estimation in applications where the sensors are subject to sudden and unpredictable variation in quality. For an illustrative trajectory, the performance of the algorithm is contrasted with an extended Kalman filter (EKF) and an image-enhanced algorithm based upon the nominal sensors.

State estimation with biased observations

State estimation is difficult when the system has multiple modes of operation. Modal transitions create discontinuities in the reference point for the local state variables. The uncertain reference point increases the ambiguity in the state measurement. The paper presents an estimation algorithm that can be used in multimodal applications. The algorithm is shown to be superior to the Kalman filter when the state measurement is contaminated with a mode dependent offset. Despite the uncertain reference point in the observation, good estimates of the underlying entire state processes can be generated.

Tracking ground vehicles on a grid

Hybrid models have proven useful for tracking targets with multiple motion modes. Most emphasis in the literature has been devoted to aircraft which transition from constant velocity motion to constant (or nearly constant) turns and back. Ground targets motions have received less attention despite similarities with aircraft. This paper presents a study of the ground-tracking problem using the Gaussian wavelet estimator as the basic algorithm. The sensor suite contains a matrix of range-bearing sensors of quality that is strongly range dependent. There also may be an acoustic sensor which provides an auxiliary speed measurement. It is shown that the high degree of partitioning of the kinematic state space provided by the algorithm is useful in this application.