PhD Thesis: Gabriel Lellouch


Adobe-PDF-downloadLellouch, Gabriel. Waveform Design and Processing Techniques in OFDM Radar. PhD Thesis. Department of Electrical Engineering, University of Cape Town, 2015.



With the advent of powerful digital hardware, software defined radio and radar have become an active area of research and development. This in turn has given rise to many new research directions in the radar community, which were previously not comprehensible. One such direction is the recently investigated OFDM radar, which uses OFDM waveforms instead of the classic linear frequency modulated waveforms. Being a wideband signal, the OFDM symbol offers spectral effiency along with improved range resolution, two enticing characteristics for radar. Historically a communication signal, OFDM is a special form of multicarrier modulation, where a single data stream is transmitted over a number of lower rate carriers. The information is conveyed via sets of complex phase codes modulating the phase of the carriers. At the receiver, a demodulation stage estimates the transmitted phase codes and the information in the form of binary words is finally retrieved. In radar, the primary goal is to detect the presence of targets and possibly estimate some of their features through measurable quantities, e.g. range, Doppler, etc.. Yet, being a young waveform in radar, more understanding is required to turn it into a standard radar waveform.

Our goal, with this thesis, is to mature our comprehension of OFDM for radar and contribute to the realm of OFDM radar.

First, we develop two processing alternatives for the case of a train of wideband OFDM pulses. In this, our first so-called time domain solution consists in applying a matched filter to compress the received echoes in the fast time before applying a fast Fourier transform in the slow time to form the range Doppler image. We motivate this approach after demonstrating that short OFDM pulses are Doppler tolerant. The merit of this approach is to conserve existing radar architectures while operating OFDM waveforms. The second so-called frequency domain solution that we propose is inspired from communication engineering research since the received echoes are tumbled in the frequency domain. After several manipulations, the range Doppler image is formed. We explain how this approach allows to retrieve an estimate of the unambiguous radial velocity, and propose two methods for that. The first method requires the use of identical sequence
(IS) for the phase codes and is, as such, binding, while the other method works irrespective of the phase codes. Like the previous technique, this processing solution accommodates high Doppler frequencies and the degradation in the range Doppler image is negligible provided that the spacing between consecutive subcarriers is sufficient. Unfortunately, it suffers from the issue of intersymbol interference (ISI). After observing that both solutions provide the same processing gain, we clarify the constraints that shall apply to the OFDM signals in either of these solutions. In the first solution, special care has been employed to design OFDM pulses with low peak-to-mean power ratio (PMEPR) and low sidelobe level in the autocorrelation function. In the second solution, on the other hand, only the constraint of low PMEPR applies since the sidelobes of the scatterer characteristic function in the range Doppler image are Fourier based.

Then, we develop a waveform-processing concept for OFDM based stepped frequency waveforms. This approach is intended for high resolution radar with improved low probability of detection (LPD) characteristics, as we propose to employ a frequency hopping scheme from pulse to pulse other than the conventional linear one. In the same way we treated our second alternative earlier, we derive our high range resolution processing in matrix terms and assess the degradation caused by high Doppler on the range profile. We propose using a bank of range migration filters to retrieve the radial velocity of the scatterer and realise that the issue of classical ambiguity in Doppler can be alleviated provided that the relative bandwidth, i.e. the total bandwidth covered by the train of pulses divided by the carrier frequency, is chosen carefully. After discussing a deterministic artefact caused by frequency hopping and the means to reduce it at the waveform design or processing level, we discuss the benefit offered by our concept in comparison to other standard wideband methods and emphasize on its LPD characteristics at the waveform and pulse level.

In our subsequent analysis, we investigate genetic algorithm (GA) based techniques to finetune OFDM pulses in terms of radar requirements viz., low PMEPR only or low PMEPR and low sidelobe level together, as evoked earlier. To motivate the use of genetic algorithms, we establish that existing techniques are not exible in terms of the OFDM structure (the assumption that all carriers are present is always made). Besides, the use of advanced objective functions suited to particular configurations (e.g. low sidelobe level in proximity of the main autocorrelation peak) as well as the combination of multiple objective functions can be done elegantly with GA based techniques. To justify that solely phase codes are used for our optimisation(s), we stress that the weights applied to the carriers composing the OFDM signal can be spared to cope with other radar related challenges and we give an example with a case of enhanced detection.

Next, we develop a technique where we exploit the instantaneous wideband transmission to characterise the type of the canonical scatterers that compose a target. Our idea is based on the well-established results from the geometrical theory of diffraction (GTD), where the scattered energy varies with frequency. We present the problem related to ISI, stress the need to design the transmitted pulse so as to reduce this risk and suggest having prior knowledge over the scatterers relative positions. Subsequently, we develop a performance analysis to assess the behaviour of our technique in the presence of additive white Gaussian noise (AWGN). Then, we demonstrate the merit of integrating over several pulses to improve the characterisation rate of the scatterers.

Because the scattering centres of a target resonate variably at different frequencies, frequency diversity is another enticing property which can be used to enhance the sensing performance. Here, we exploit this element of diversity to improve the classification function. We develop a technique where the classification takes place at the waveform design when few targets are present. In our case study, we have three simple targets. Each is composed of perfectly electrically conducting spheres for which we have exact models of the scattered field. We develop a GA based search to find optimal OFDM symbols that best discriminate one target against any other. Thereafter, the OFDM pulse used for probing the target in the scene is constructed by stacking the resulting symbols in time. After discussing the problem of finding the best frequency window to sense the target, we develop a performance analysis where our figure of merit is the overall probability of correct classification. Again, we prove the merit of integrating over several pulses to reach classification rates above 95%.

In turn, this study opens onto new challenges in the realm of OFDM radar. We leave for future research the demonstration of the practical applicability of our novel concepts and mention manifold research axes, viz., a signal processing axis that would include methods to cope with intersymbol interference, range migration issues, methods to raise the ambiguity in Doppler when several echoes from distinct scatterers overlap in the case of our frequency domain processing solutions; an algorithmic axis that would concern the heuristic techniques employed in the design of our OFDM pulses. We foresee that further tuning might help speeding up our GA based algorithms and we expect that constrained multiobjective optimisation GA (MOO-GA) based techniques shall benefit the OFDM pulse design problem in radar. A system design axis that would account for the hardware components’ behaviours, when possible, directly at the waveform design stage and would include implementation of the OFDM radar system.