Multi-Modal Sensor System Integrating COTS Technology for Surveillance and Tracking Robert Kozma a , Lan Wang a , Khan Iftekharuddin a a University of Memphis, Memphis, TN 38152, USA rkozma@memphis.edu Ernest McCracken a , Mohammed Khan a , Khandekar Islam a , Rustu M. Demirer b b Istanbul Kultur University, Istanbul, Turkey Abstract— Abstract— The feasibility of low-cost, Commercial Off-The-Shelf (COTS) sensor nodes is studied in a distributed network, aiming at dynamic surveillance and tracking of ground targets. Data acquisition by the low-cost (< $50 US) miniature radar is described. We demonstrate the detection, ranging and velocity estimation capabilities of the mini-radar, and compare results to simulations. Furthermore, we integrate the radar output with supplementary sensor modalities, such as acoustic and vibration transducers, and infrared sensors. The method provides innovative solutions for detecting, identifying, and tracking vehicles and dismounts over a wide area in noisy conditions. This study presents a step towards distributed intelligent decision support and demonstrates effectiveness of small cheap sensors in identifying unique but similar events. Our work supports a pervasive sensor implementation relevant to the affordable open system architecture attribute of the U.S. Air Force Layered Sensing paradigm. 1 I. INTRODUCTION N autonomous sensor network is a collection of sensor nodes with limited processing, power, and communication capabilities that monitor a real world environment through differing modalities. The nodes gather information about the local environment, preprocess the data, and transmit the output via wireless channels to a base station. The base station may broadcast commands to all or some of the sensor nodes in the network. Using intelligent framework design, the network can support decision making and provide the capability to detect, track, and identify targets over a wide area. Sensor networks are often composed of sensor nodes mounted on autonomous vehicles, sensor nodes on the ground, and remote base stations. In actual deployment environments, a node may not be able to reach all the other nodes within its physical wireless transmission range, i.e. not all the nodes are within one hop of the others. Challenging research issues include dynamic network partition due to 1 This research has been supported in part by a research grant of the FedEx Institute of Technology, the University of Memphis. The work by one of the authors (R.K.) has been supported by Air Force Office of Scientific Research. (AFOSR/NL), Dr. Jun Zhang. node mobility, node failures, wireless channel interference/ fading, and network congestion [1]. We need to assure reliable communication among the nodes and with a remote base station, despite these various adverse factors. Given the harsh operating environment of the proposed system, our foremost design goal should be robustness. A relevant area of research is delay tolerant networks [2]. Another requirement is to minimize the target detection delay and maximize classification accuracy, while minimizing the consumption of network resources (bandwidth, memory, and power). In this work we introduce a sensory network consisting of several COTS sensors including radar, infrared, acoustic, and magnetic nodes, that can be used as an integrated surveillance and sensing system. Short-rage radar detectors has been used in commercial applications, including traffic management, and proximity sensing [3,4]. We describe the acquisition properties and implementation of a low-cost (< $50 US) miniature radar. We then demonstrate the detection, ranging and velocity estimation capabilities of the mini-radar. We discuss the integration of the radar and other sensor data and the possibilities of developing a robust pervasive sensor system using multi-modal technology, relevant to the affordable open system architecture attribute of the U.S. Air Force Layered Sensing paradigm [5]. II. MINIATURE MULTI-FUNCTION RADAR This section describes the small, low-cost, COTS K-band radars that are used on the sensor nodes to help detect and characterize targets. We also describe the COTS hardware that combines the various sensor modalities into a single, autonomous node. The RF transceivers are manufactured by M/A-COM (Tyco Electronics) model MACS-007802- 0M1RSV, and are used primarily for automotive applications, e.g., front and rear-end collision detection, in ground speed measurements, and as motion detectors, e.g., automatic door openers [6]. The radar utilizes a Gunn diode oscillator and transmits a continuous wave at 24.125 GHz. It also has 0.3 GHz of bandwidth that can be controlled by applying an external voltage, making it is capable of estimating target A