Collaborative In-Network Processing for Target Tracking
- PDF / 1,438,129 Bytes
- 14 Pages / 600 x 792 pts Page_size
- 43 Downloads / 278 Views
Collaborative In-Network Processing for Target Tracking Juan Liu Palo Alto Research Center, 3333 Coyote Hill Road, Palo Alto, CA 94304, USA Email: [email protected]
James Reich Palo Alto Research Center, 3333 Coyote Hill Road, Palo Alto, CA 94304, USA Email: [email protected]
Feng Zhao Palo Alto Research Center, 3333 Coyote Hill Road, Palo Alto, CA 94304, USA Email: [email protected] Received 21 December 2001 and in revised form 4 October 2002 This paper presents a class of signal processing techniques for collaborative signal processing in ad hoc sensor networks, focusing on a vehicle tracking application. In particular, we study two types of commonly used sensors—acoustic-amplitude sensors for target distance estimation and direction-of-arrival sensors for bearing estimation—and investigate how networks of such sensors can collaborate to extract useful information with minimal resource usage. The information-driven sensor collaboration has several advantages: tracking is distributed, and the network is energy-efficient, activated only on a when-needed basis. We demonstrate the effectiveness of the approach to target tracking using both simulation and field data. Keywords and phrases: sensor network, target tracking, distributed processing, Bayesian filtering, beamforming, mutual information.
1.
INTRODUCTION
Sensors of various types have already become ubiquitous in modern life, from infrared motion detectors in our light switches to silicon accelerometers in the bumpers of our cars. As the cost of the sensors comes down rapidly due to advances in MEMS fabrication and because these sensors increasingly acquire networking and local processing capabilities, new types of software applications become possible, distributed among these everyday devices and performing functions previously impossible for any of the devices independently. Enabling such functionality without overtaxing the resources of the existing devices, especially when these devices are untethered and running on batteries, may require us to rethink some important aspects of how sensing systems are designed. 1.1. Advantages of distributed sensor networks There are a number of reasons why networked sensors have a significant edge over existing, more centralized sensing platforms. An ad hoc sensor network can be flexibly deployed in an area where there is no a priori sensing infrastructure. Coverage of a large area is important for tracking events of a significant spatial extent as in tracking events of a significant
spatial extent, as in tracking a large number of events simultaneously, or for tracking dynamic events traversing the sensing ranges of many individual sensors, as in tracking a moving vehicle. In cases of tracking low-observable phenomena, such as a person walking in an obstacle field in an urban environment or a stealthy military vehicle, the signal-to-noise ratio (SNR) of data collected from a central location may be unacceptable. As sensor density increases, the mean distance from the nearest sensor to a target decreases and the SNR received at th
Data Loading...
