Orthogonal frequency division multiplexing multiple-input multiple-output automotive radar with novel signal processing algorithms

Abstract

Advanced driver assistance systems that actively assist the driver based on environment perception achieved significant advances in recent years. Along with this development, autonomous driving became a major research topic that aims ultimately at development of fully automated, driverless vehicles. Since such applications rely on environment perception, their ever increasing sophistication imposes growing demands on environmental sensors. Specifically, the need for reliable environment sensing necessitates the development of more sophisticated, high-performance radar sensors. A further vital challenge in terms of increased radar interference arises with the growing market penetration of the vehicular radar technology. To address these challenges, in many respects novel approaches and radar concepts are required. As the modulation is one of the key factors determining the radar performance, the research of new modulation schemes for automotive radar becomes essential. A topic that emerged in the last years is the radar operating with digitally generated waveforms based on orthogonal frequency division multiplexing (OFDM). Initially, the use of OFDM for radar was motivated by the combination of radar with communication via modulation of the radar waveform with communication data. Some subsequent works studied the use of OFDM as a modulation scheme in many different radar applications - from adaptive radar processing to synthetic aperture radar. This suggests that the flexibility provided by OFDM based digital generation of radar waveforms can potentially enable novel radar concepts that are well suited for future automotive radar systems. This thesis aims to explore the perspectives of OFDM as a modulation scheme for high-performance, robust and adaptive automotive radar. To this end, novel signal processing algorithms and OFDM based radar concepts are introduced in this work. The main focus of the thesis is on high-end automotive radar applications, while the applicability for real time implementation is of primary concern. The first part of this thesis focuses on signal processing algorithms for distance-velocity estimation. As a foundation for the algorithms presented in this thesis, a novel and rigorous signal model for OFDM radar is introduced. Based on this signal model, the limitations of the state-of-the-art OFDM radar signal processing are pointed out. To overcome these limitations, we propose two novel signal processing algorithms that build upon the conventional processing and extend it by more sophisticated modeling of the radar signal. The first method named all-cell Doppler compensation (ACDC) overcomes the Doppler sensitivity problem of OFDM radar. The core idea of this algorithm is the scenario-independent correction of Doppler shifts for the entire measurement signal. Since Doppler effect is a major concern for OFDM radar and influences the radar parametrization, its complete compensation opens new perspectives for OFDM radar. It not only achieves an improved, Doppler-independent performance, it also enables more favorable system parametrization. The second distance-velocity estimation algorithm introduced in this thesis addresses the issue of range and Doppler frequency migration due to the target’s motion during the measurement. For the conventional radar signal processing, these migration effects set an upper limit on the simultaneously achievable distance and velocity resolution. The proposed method named all-cell migration compensation (ACMC) extends the underlying OFDM radar signal model to account for the target motion. As a result, the effect of migration is compensated implicitly for the entire radar measurement, which leads to an improved distance and velocity resolution. Simulations show the effectiveness of the proposed algorithms in overcoming the two major limitations of the conventional OFDM radar signal processing. As multiple-input multiple-output (MIMO) radar is a well-established technology for improving the direction-of-arrival (DOA) estimation, the second part of this work studies the multiplexing methods for OFDM radar that enable simultaneous use of multiple transmit antennas for MIMO radar processing. After discussing the drawbacks of known multiplexing methods, we introduce two advanced multiplexing schemes for OFDM-MIMO radar based on non-equidistant interleaving of OFDM subcarriers. These multiplexing approaches exploit the multicarrier structure of OFDM for generation of orthogonal waveforms that enable a simultaneous operation of multiple MIMO channels occupying the same bandwidth. The primary advantage of these methods is that despite multiplexing they maintain all original radar parameters (resolution and unambiguous range in distance and velocity) for each individual MIMO channel. To obtain favorable interleaving patterns with low sidelobes, we propose an optimization approach based on genetic algorithms. Furthermore, to overcome the drawback of increased sidelobes due to subcarrier interleaving, we study the applicability of sparse processing methods for the distance-velocity estimation from measurements of non-equidistantly interleaved OFDM-MIMO radar. We introduce a novel sparsity based frequency estimation algorithm designed for this purpose. The third topic addressed in this work is the robustness of OFDM radar to interference from other radar sensors. In this part of the work we study the interference robustness of OFDM radar and propose novel interference mitigation techniques. The first interference suppression algorithm we introduce exploits the robustness of OFDM to narrowband interference by dropping subcarriers strongly corrupted by interference from evaluation. To avoid increase of sidelobes due to missing subcarriers, their values are reconstructed from the neighboring ones based on linear prediction methods. As a further measure for increasing the interference robustness in a more universal manner, we propose the extension of OFDM radar with cognitive features. We introduce the general concept of cognitive radar that is capable of adapting to the current spectral situation for avoiding interference. Our work focuses mainly on waveform adaptation techniques; we propose adaptation methods that allow dynamic interference avoidance without affecting adversely the estimation performance. The final part of this work focuses on prototypical implementation of OFDM-MIMO radar. With the constructed prototype, the feasibility of OFDM for high-performance radar applications is demonstrated. Furthermore, based on this radar prototype the algorithms presented in this thesis are validated experimentally. The measurements confirm the applicability of the proposed algorithms and concepts for real world automotive radar implementations.