Steve Tanner

Real time target tracking with binary sensor networks and parallel computing

A parallel real time data fusion and target tracking algorithm for very large binary sensor networks is presented. A binary sensor can give an on or off signal to indicate the presence or absence of targets within its range, but it cannot tell how many targets are present, where the targets are, how fast they are moving, or which direction they are heading. In order to detect and track targets using these sensors, it is necessary to fuse information from more than one sensor. A parallel data fusion process based on simulated annealing is used to identify and locate targets. Processing is performed on a commodity Linux cluster with communication between nodes facilitated by the Message Passing Interface (MPI). The fusion and tracking algorithm is tested with a wide variety of sensor network parameters using target track data from a theater level air combat simulation. It is demonstrated that very accurate target detection and localization are possible even though the binary sensors themselves provide little information and have high error rates. Real time tracking is performed on a network with 2.5 million sensors on a commodity cluster with only 50 processors.

On-board Data Mining

Networks of remote sensors are becoming more common as technology improves and costs decline. In the past, a remote sensor was usually a device that collected data to be retrieved at a later time by some other mechanism. This collected data were usually processed well after the fact at a computer greatly removed from the in situ sensing location. This has begun to change as sensor technology, on-board processing, and network communication capabilities have increased and their prices have dropped. There has been an explosion in the number of sensors and sensing devices, not just around the world, but literally throughout the solar system. These sensors are not only becoming vastly more sophisticated, accurate, and detailed in the data they gather but they are also becoming cheaper, lighter, and smaller. At the same time, engineers have developed improved methods to embed computing systems, memory, storage, and communication capabilities into the platforms that host these sensors. Now, it is not unusual to see large networks of sensors working in cooperation with one another. Nor does it seem strange to see the autonomous operation of sensorbased systems, from space-based satellites to smart vacuum cleaners that keep our homes clean and robotic toys that help to entertain and educate our children. But access to sensor data and computing power is only part of the story. For all the power of these systems, there are still substantial limits to what they can accomplish. These include the well-known limits to current Artificial Intelligence capabilities and our limited ability to program the abstract concepts, goals, and improvisation needed for fully autonomous systems. But it also includes much more basic engineering problems such as lack of adequate power, communications bandwidth, and memory, as well as problems with the geolocation and real-time georeferencing required to integrate data from multiple sensors to be used together.

Real-time target tracking simulations in large disparate sensor networks

Real-time target tracking in large disparate sensor networks has been simulated with a parallelized search based data fusion algorithm using a simulated annealing approach. The networks are composed of large numbers of low fidelity binary and bearing-only sensors, and small numbers of high fidelity position sensors over a large region. The primitive sensors provide limited information, not sufficient to locate targets; the position sensors can report both range and direction of the targets. Target positions are determined through fusing information from all types of sensors. A score function, which takes into account the fidelity of sensors of different types, is defined and used as the evaluation function for the optimization search. The fusion algorithm is parallelized using spatial decomposition so that the fusion process can finish before the arrival of the next set of sensor data. A series of target tracking simulations are performed on a Linux cluster with communication between nodes facilitated by the Message Passing Interface (MPI). The probability of detection (POD), false alarm rate (FAR), and average deviation (AVD) are used to evaluate the network performance. The input target information used for all the simulations is a set of target track data created from a theater level air combat simulation.

A heterogeneous sensor network simulation system with integrated terrain data for real-time target detection in 3D space

Large scale sensor networks composed of many low-cost small sensors networked together with a small number of high fidelity position sensors can provide a robust, fast and accurate air defense and warning system. The team has been developing simulations of such large networks, and is now adding terrain data in an effort to provide more realistic analysis of the approach. This work, a heterogeneous sensor network simulation system with integrated terrain data for real-time target detection in a three-dimensional environment is presented. The sensor network can be composed of large numbers of low fidelity binary and bearing-only sensors, and small numbers of high fidelity position sensors, such as radars. The binary and bearing-only sensors are randomly distributed over a large geographic region; while the position sensors are distributed evenly. The elevations of the sensors are determined through the use of DTED Level 0 dataset. The targets are located through fusing measurement information from all types of sensors modeled by the simulation. The network simulation utilizes the same search-based optimization algorithm as in our previous two-dimensional sensor network simulation with some significant modifications. The fusion algorithm is parallelized using spatial decomposition approach: the entire surveillance area is divided into small regions and each region is assigned to one compute node. Each node processes sensor measurements and terrain data only for the assigned sub region. A master process combines the information from all the compute nodes to get the overall network state. The simulation results have indicated that the distributed fusion algorithm is efficient enough so that an optimal solution can be reached before the arrival of the next sensor data with a reasonable time interval, and real-time target detection can be achieved. The simulation was performed on a Linux cluster with communication between nodes facilitated by the Message Passing Interface (MPI). The input target information for the simulations is a set of modified target track data generated from a realistic theater level air combat simulation. The probability of detection (POD), false alarm rate (FAR), and average deviation (AVD) are used in evaluating the network performance.