Advanced TOA Estimation for Multipath Channels
Rabih Chrabieh, Peter Bagnall, Serdar Sezginer, Nestwave, France
Location: Atrium Ballroom
Date/Time: Tuesday, Apr. 21, 1:50 p.m.
As the Internet of Things (IoT) continues its rapid roll-out across much of the globe, geolocation has become among the IoT’s most-used features. However, accurate positioning with cost- and power-efficient solutions are still being investigated intensively in order to address challenging use cases. In particular, estimation of time of arrival (TOA) in multipath environments for terrestrial or GNSS systems limits performance and can benefit from special processing with advanced algorithms.
In a typical TOA estimation problem, the channel response consists of delayed taps (i.e. multipaths or a sequence of Dirac impulses) convolved with the impulse response of some low pass filter. Given a limited bandwidth, each delayed tap is convolved with a sinc or similar low-pass waveform, and summed with the rest. The received waveform is convoluted, taps interfere with each other and it is often difficult to separate them.
The impulse response shape is usually a sinc, a raised cosine, or some relatively slowly decaying waveform symmetrical around the origin. The reason for the symmetry is that a frequency domain matched filter, necessary to optimize symbol detection for data transmission, creates a zero-phase signal symmetrical in time domain.
However, for estimating the TOA where the information is typically carried by the first tap only, a matched filter is not ideal. In particular, using a non-symmetrical, near-causal filter (near zero energy before the origin) ensures that the first tap is less interfered by subsequent taps.
A filter with characteristics varying between a matched filter and a near-causal filter can be used depending on signal to noise ratio (SNR). At very low SNR, a matched filter produces better results, while at very high SNR a near-causal filter can better isolate the first tap from interfering taps. We found even better performance at medium to high SNR through the development of a “TOA matched filter” which took into account the expected channel’s delay profile
In this paper, we introduce advanced methods for designing near-causal filters or TOA-matched filters. The filters are generally efficient and robust to apply and remove a significant amount of delayed multipath interference. They are especially powerful and robust when multipath interference is dense, and work well in NLOS scenarios: the filter extracts the useful signal from underneath strong multipath. The filter design relies on some modeling of the channel's power delay profile but it can tolerate wide variations in the channel.
Post application of the filter, having reduced relatively distant multi-path components, we then propose a "reduced complexity" Maximum Likelihood estimation of the TOA using 2 or more optimally located points to extract the first path from nearby multipath.
The proposed estimation procedure is applicable to all the wireless systems such as 4G, 5G, GNSS, WiFi, UWB, etc, which suffer significantly from multipath interference. This opens the door for optimized hybrid solutions which allow cost-effective and ultra low-power solutions for accurate estimations. Results show that, in such systems, significant improvements are obtained with respect to the existing solutions both in terms of estimation performance and implementation aspects.