Wireless Sensor Network with Geolocation

June 1, 2009

Maintaining situational/positional awareness in indoor and urban environments is difficult because buildings, walls, and other obstacles obstruct vision and RF propagation. Propagation channels are typically characterized by direct line-of-sight (DLOS) and non-line-of-sight (NLOS) performance. As more obstruction is added to the propagation channel, the DLOS path amplitude decreases relative to the NLOS path due to multipath and shadow fading. Thus, the DLOS path, which is the key enabling parameter for geolocation, is harder to detect. The challenge is to develop geolocation algorithms that overcome the DLOS obstructions, and integrate the algorithms into a communication network that is easily deployed, low cost, and provides accurate and reliable location information.

Figure 1. A comparison of the time domain waveform for bandwidths of 3 GHz, 500 MHz, and 10 MHz.

The geolocation requirements include lightweight, low-power, portable devices that can be rapidly deployed in an ad-hoc network. Each node will provide high-resolution, near-real-time geolocation for situational awareness in urban/ indoor operations. The devices must provide robust data communication for exchange of information and operate in the presence of undesired interference. The military devices must also provide low probability of detection or interception by the enemy.

Project Objectives

In a project with the Army Advanced Research Lab, the key objectives were:

Establish scientific foundation for evaluation of indoor and urban geolocation systems.
Develop innovative algorithms for ultra-wideband (UWB) geolocation.
Define suitable architecture for realtime implementation.
Develop working prototype radios and demonstrate real-time experiment.
Provide user with robust real-time method to determine accurate position estimation.
Develop solutions for DoD, homeland security, and public safety.

A key task of the geolocation project is to collect path data to establish an empirical model specific to detection of the DLOS path for indoor-to-indoor and outdoor-to-indoor environments. This model is used to develop and optimize signal processing algorithms and a system architecture that provides real-time geolocation capability.

Technical Approach

The geolocation design uses an ultra-wideband waveform based on the WiMedia group MB-OFDM standard. The 528-MHz bandwidth waveform provides 200 bits per symbol to achieve very high data rates at short range. An algorithm variation was added to achieve ranges of greater than 250 meters and provide range versus data rate flexibility from 200 Mbps @ 10 meters to 300 kilobits/second @ 250 meters. The waveform is designed for high-multi-path indoor environments and supports tailoring of the RF spectrum to mitigate RF interference and coexistence issues.

Data Collection

The key to developing a geolocation waveform is to define a realistic path model. A data collection system was defined to measure the im pulse response of the propagation channel over an RF spectrum of 3-6 GHz using a network analyzer.

Figure 2. The proof-of-concept hardware assembly is approximately 12 x 12.5 x 3.5" including the electronics and battery.

The s-parameters obtained from the network analyzer can be applied to multiple bandwidths using the Agilent ADS and MATLAB to compare various processing algorithms. Hundreds of measurements have been collected for indoor-to-indoor and outdoor-to-indoor scenarios for multiple building constructions to develop empirical path models. These models are especially appropriate for analyzing the performance of geolocation algorithms based on time-of-arrival (TOA).

Figure 1 shows a comparison of the time domain waveform for bandwidths of 3 GHz, 500 MHz, and 10MHz. Both the 3- GHz and 500-MHz bandwidths allow separation of the direct path from multipath signals. However, the 10-MHz bandwidth provides no separation, which prevents a good TOA measurement. The TOA measurement is based on Two-Way Time Transfer (TWTT), which is well suited for ad-hoc mesh deployment since there is no infrastructure and precise synchronization to a common clock is not required.

The TWTT calculation is also useful for network time-synchronization, frequency-offset error measurement, and error correction. Based on the data collection, two distance-power relationships have been developed for geolocation. Wireless Sensor Network with Geolocation The first is a one-piece line based on a continuous path from transmitter to receiver. The second is a two-piece line based on discontinuous path from transmitter to receiver where the breakpoint in the curve occurs at the distance of the first obstruction. UWB channel characteristics from the data collection of indoor-to-indoor and outdoor-to-indoor conditions include the following:

DLOS has high path loss with power exponent of 6 to 10 typical at distances >10m.
A dense multi-path impulse response of up to several hundred nanoseconds is typical.
There can be up to 100 nanoseconds between DLOS path and strongest path (more for outdoors).
Typical power difference between DLOS path and strongest path is 26 dB with up to 40 dB observed.

Architecture and Platform

The geolocation network concept is an ad hoc mesh that includes three types of nodes: Mobile, Gateway, and Reference Nodes. Mobile nodes are carried or worn by soldiers and first responders. They are small, lightweight, and battery operated, and can interface with biometric sensors as well as other equipment such as Future Force Warrior multi-function display/touchpad. Gate - way nodes perform inter-network routing and incorporate an interface to link the mesh network to a wide-area command and control network. Reference nodes are used to orient, anchor, and/or extend the range of the network. These nodes are packaged for ease of deployment, are battery operated, and include a GPS receiver to provide a network reference. Strategic placement of reference nodes is important to good geolocation performance. In the event that GPS is not available, the system provides relative location of each unit in the network.

The geolocation waveform is similar to the MBOFDM sync header, which is processed using a matched filter and integrator based on the extended-range algorithm. Further processing is required by the ranging algorithm to measure time delay and signal strength of the DLOS signal and peak signal. This information is passed to the ranging refinement function and positioning filter. The signal processing provides rapid sync acquisition of <100µs. Ranging distance varies as a function of path conditions. Distance of 50-100 meters has been demonstrated for a UWB waveform operating at FCC emissions for NLOS conditions.

A number of positioning algorithms were evaluated including non-cooperative algorithms and cooperative algorithms. The non-cooperative algorithms included Least Square, Weighted Least Square, Residual Weighted Least Square, and Davidon. The cooperative algorithms included Savarese and an algorithm called Cooperative Localization with Optimum Quality of Estimate (CLOQ). In all scenarios, the cooperative algorithms provided better overall performance due to the exchange of positioning information between nodes. The CLOQ algorithm provided the best performance because it includes a quality of estimate (QoE) parameter based on the relative signal strength of each RF link.

The CLOQ positioning sequence is as follows. On the first iteration after startup, the reference nodes broadcast and the mobile nodes listen. Mobile nodes with an adequate number of references calculate their location and QoE. Each node with estimates of their location then broadcasts their location and QoE. Nodes with best QoE establish themselves as anchors. On the next iteration, the newly elected anchors plus original anchors are used for selection of another set of anchors. The positioning algorithm repeats until all nodes become anchors and estimate their QoE. The average position error varied from 1.5 meters for 5 mobile nodes to <0.5 meter for 40 nodes.

The platform architecture for the geolocation node includes a wideband transceiver that supports the 528-MHz bandwidth signal, dual A/D and dual D/A, two Altera FPGAs for transmit and receive, and an Intel X-Scale general-purpose processor. The proof-of-concept hardware assembly is approximately 12 × 12.5 × 3.5" including the electronics and battery (see Figure 2). The user interface provides node locations on a map background or other backgrounds such as concentric circles with the user node in the center.

The multimode radio design will be a 4" diameter form factor. The design is based on an optimization of signal processing algorithms developed from an analysis of empirical geolocation path models. The UWB platform architecture supports the geolocation algorithms for widely distributed localization of radio nodes. Each user node supports a display device with user interface for situation awareness. The design minimizes data throughput and power consumption, and tolerates implementation limitations such as RF frequency error, phase noise, quantization noise, and sampling phase jitter.

This article was written by James Silverstrim and Roderick Passmore of Innovative Wireless Technologies in Forest, VA; Dr. Kaveh Pahlavan of Worcester Polytechnic Institute in Worchester, MA; and Dr. Brian Sadler of the US Army Advanced Research Laboratory in Adelphi, MD. For more information, Click Here