Supervisor: A/Prof. Tomonari Furukawa
PhD Student: Lin Chi Mak
 

Contact
Lin Chi (Edward) Mak

School of Mechanical and Manufacturing Engineering,
The University of New South Wales
NSW, 2052 Australia
Telephone: +61-2-9385-6091
Fax: +61-2-9663-1222

Background

Without knowing the position of a Unmanned Ground Vehicle (UGV), many indoor autonomous tasks, such as surveillance and Urban Search And Rescue (USAR) are not possible. In an indoor environment, where GPS data is unavailable or unreliable, the presence of obstacles can result in erroneous positioning data in line-of-sight localisation systems. Various indoor localisation systems, such as ultrasonic and radio signal strength (RSS), have been developed, but their positioning accuracies are still not good enough (>1m) for indoor ground vehicles to navigate under Non-Line-of-Sight (NLoS) conditions (There is a blockage between the target and the ground stations).

Objective

This project is aimed at developing a novel NLoS mitigation technique and its system for a ground vehicle. Such a system should provide accurate trajectory data for UGV (<0.5m) to navigate in indoors even in NLOS conditions.

Approach

In this project, a ToA-based NLoS technique and its system have been proposed. The proposed technique first utilizes the low-frequency sound to minimize the NLoS errors (positive bias error in ToA). Since the high diffraction nature of low-frequency sound, such a signal can arrive the target through the shortest diffracted path, giving smallest NLoS error. Then, the true ToAs are estimated by the modeled map without notable bias errors, and then compared with the measured to update the target position. The strength of the proposed technique is that the positioning accuracy is not corrupted by any small or known obstacle.

In the developed system, a microphone is mounted on a UGV and 3-4 speakers are located at known positions as base stations. The received signals are sent to and analysed by a computer. The Signal-to-Noise Ratio (SNR) of the received signal is enhanced by applying narrow bandpass filter. The ToA of the first arrival signal is estimated by finding the first peak exceeding the set threshold and used it for localization of the UGV. More details can be found in the related publication.

Results

The technique and system have been built and tested successfully. Fig.2 and Fig.3 shows estimated path by proposed techniques and other 3 techniques in literature in simulation and experiment respectively. In both cases, the estimated paths by the proposed technique are the closest to the true paths. The experimental results show 300%-400% improvement by the proposed technique over 3 existing methods. Table I first shows the positioning accuracies under different noise level, while Table II list the properties of the developed system. In conclusion, this project has developed a cost-effective and accurate NLoS localisation system.

On-going Research

Currently we are working on a new NLOS localization technique which uses Frequency Power Delay Profile (FPDP) of the first-arrival signal for NLoS localization problem. Since the first-arrival signal is always from the shortest diffracted path, the signal loss of the first-arrival signal in NLoS conditions is dependent on the frequency of the signal. Based on our proposed wave propagation model in a known map, the Frequency Power Delay Profile Fingerprint (FPFPF) of the first-arrival signals at different positions can be estimated without any pre-measurement of the FPDP.

The measured FPDP is computed by taking the correlation of the emitted signal by a sound source, which is our target, and the received signal by a microphone at known position. The measured FPDP is then compared with the estimated FPDPF to give the position of the target. In our developing system, the signal from the shortest path is enhanced by using a directional microphone, giving higher Signal-to-Noise Ratio (SNR). In addition, since signals from multi-path are ignored, the performance of the system becomes more reliable. The system is insensitive to most environment noise which is usually not correlated to the emitted signal. In the recent results, the accuracy of the measured shortest path distance in NLOS conditions is around 3cm. It is expected that the system will give less than 10cm positioning errors with 3 or more directional microphones.

Publications

1. Lin Chi Mak and Tomonari Furukawa, “Non-Line-of-Sight Localization of a Controlled Sound Source,” 2009 IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM09), Singapore, Singapore, July 14-17, 2009, 7 pages, accepted.

2. Lin Chi Mak and Tomonari Furukawa, "A Time-of-Arrival-Based Positioning Technique with Non-Line-of-Sight Mitigation Using Low-frequency Sound", Advanced Robotics, Vol. 22, no. 5, pp. 507-526, 2008.

3. Lin-Chi Mak and Tomonari Furukawa, "A ToA-based Approach to NLOS Localization Using Low-frequency Sound", 2006 Australian Conference on Robotics and Automation (ACRA ‘06), CD-ROM, Auckland, Dec 6-8, 2006.