Automotive radar is a key element in Advanced Driver Assistance Systems (ADAS). With the growth of Automotive industry, there is a high demand for the sensors used in assisting systems. As the number of sensors increases, probability that these systems being in close proximity will also increase. This will lead to situations where in multiple radar sensors will be operating in close proximity, this might lead to a sub-optimal performance of our radar system. Currently FMCW radar systems are most prevalent in the automotive radar market. For FMCW systems, interference mitigation techniques exist in time domain, frequency domain, polarization domain and etc. Most of these techniques insiston detecting and identifying the interference before mitigating it. In this thesis, we consider a FMCW radar system and first develop a MATLAB model for the three most important interference scenarios namely Continuous Wave(CW), FrequencyModulatedContinuousWave (FMCW) and PhaseModulatedContinuousWave (PMCW). We propose a signal model where interference can be localized over the beat signal, then we systematically study how interference can be detected. As a result, we will detect the interference even if the power level of the interferer is lower than the power level of the received reflected signal. Post detection of interference,we suggest a technique to identify the interference by estimating the slope from the detected interference samples. Starting with a simple existing mitigation technique, we look at how to mitigate the interference and suggest enhancements that can be done for these techniques post detection and identification. Considering a worst case scenario of interference being completely in band to the transmitted signal, we propose a novel avoidance technique which will also predict the bandwidth of the interferer. As a result we will be able to shift the center frequency of the transmitter to avoid the interferer. Finally we propose another novel time domain mitigation technique where in without detecting or identifying the interferer, we will mitigate the interference and compare the gain of Signal to Interference plus Noise Ratio (SINR) achieved by using this technique.

Automotive radar is widely used for driving safety support and it is a key element of future autonomous vehicles. Radar sensors have the property that the performance is not to be affected by low vision conditions compared with camera sensors or laser based radar sensors, which makes it crucial for the autonomous driving system.

Automotive radar system utilizes millimeter-wave band to detect the range, velocity and direction of arrival (DOA). Commercial 24 GHz and 77GHz radar have been well developed in vehicle applications at present, and the future trend is 79 GHz solution with wider 4 GHz bandwidth, which has been defined by the European Commission in 2004 as the frequency allocation for automotive shortrange radar systems. The main modulation method of automotive radar is Frequency-Modulated Continous-Wave(FMCW), which holds advantages including high-resolution distance measurement, quick updating and lower sampling frequency required on hardware.


%good functions in the various weather condition.

The resolution and accuracy of DOA estimation depend on the number of used antennas and their physical size. To improve the performance advanced algorithms and antenna configurations can be used. So-called super-resolution techniques can improve the estimation performance but suffer when few measurements are available and the number of targets is unknown, and the performance will also be degraded by correlated signals.

Monopulse is an established technique for radars for precision angle estimation. It enables DOA realized in only one pulse and requires quite less computation complexity. Conventional monopulse with one transmitter and two receivers and phased array monopulse have been well developed especially in tracking radar system. MIMO-monopulse has been studied in some literature, but most of them utilize distributed MIMO. The research and application of colocated MIMO-monopulse are still in progress.

In FMCW automotive radar application, clutter may have a strong effect on DOA detection and estimation of targets with low RCS, e.g. pedestrian. With Space-Time Adaptive Processing(STAP), it is possible to suppress the clutter in the angle-Doppler domain. MIMO-monopulse basically utilizes digital beamformer to generate sum and difference channel, which also allows STAP processing to obtain adaptive sum and difference weighting. With STAP clutter can be canceled, thus false detection of MIMO-monopulse will be avoided.

In this master thesis, we propose the solution of angle estimation algorithm using MIMO-monopulse based on an actual automotive radar provided by NXP using colocated MIMO antenna. The algorithm is simulated through Matlab and verified on real experimental data. The performance of the algorithm is compared with conventional DOA algorithm and the advantages and disadvantages are analyzed. A feasible extension to STAP will be discussed to suppress the clutter for preventing false DOA estimation.
Since monopulse estimator has been approved high performance in the field of single target tracking (such as low RCS pedestrian), our proposed algorithm is also validated in single and multiple target DOA estimations, which is an attractive scenario for a potential application using MIMO-monopulse.

Compressive Sensing (CS) provides a new paradigm in data acquisition and signal processing based on the assumption of sparsity and the incoherence of the received signal. Based on that concept, many radar front-end architectures have been studied with the implementation of CS. In these architectures, less data are collected but the radar scene can be recovered through CS with high probability. All these CS front-ends have always been stated as less complicated but never evaluated. The main motivation of this thesis is to find aspects of the complexity and performance which can be used for the characterization of these front-ends like the gain in signal-to-noise ratio (SNR), number of components, power consumption etc. In this thesis, we investigate three radar front-end architectures in CS. The first two are the Multi-coset (MC) Sampling and Analog-to-Information Converter (AIC) which are widely suggested for telecommunications and radar systems. The fourth front-end is novel as it contains metamaterial surface antenna elements. The performance and the complexity of each architecture are evaluated. The performance is compared with respect to the conventional reference case of a uniform linear array of antennas.