Search

Radar sensors are crucial for environment perception of driver assistance systems as well as autonomous vehicles. With a rising number of radar sensors and the so far unregulated automotive radar frequency band, mutual interference is inevitable and must be dealt with. Algorithms and models operating on radar data are required to run the early processing steps on specialized radar sensor hardware. This specialized hardware typically has strict resource-constraints, i.e. a low memory capacity and low computational power. Convolutional Neural Network (CNN)-based approaches for denoising and interference mitigation yield promising results for radar processing in terms of performance. Regarding resource-constraints, however, CNNs typically exceed the hardware's capacities by far. In this paper we investigate quantization techniques for CNN-based denoising and interference mitigation of radar signals. We analyze the quantization of (i) weights and (ii) activations of different CNN-based model architectures. This quantization results in reduced memory requirements for model storage and during inference. We compare models with fixed and learned bit-widths and contrast two different methodologies for training quantized CNNs, i.e. the straight-through gradient estimator and training distributions over discrete weights. We illustrate the importance of structurally small real-valued base models for quantization and show that learned bit-widths yield the smallest models. We achieve a memory reduction of around 80\% compared to the real-valued baseline. Due to practical reasons, however, we recommend the use of 8 bits for weights and activations, which results in models that require only 0.2 megabytes of memory., Comment: 15 pages; published in IEEE Journal of Selected Topics in Signal Processing, Special Issue on Recent Advances in Automotive Radar Signal Processing, Volume: 15, Issue: 4, June 2021. arXiv admin note: text overlap with arXiv:2011.12706

Autonomous driving highly depends on capable sensors to perceive the environment and to deliver reliable information to the vehicles' control systems. To increase its robustness, a diversified set of sensors is used, including radar sensors. Radar is a vital contribution of sensory information, providing high resolution range as well as velocity measurements. The increased use of radar sensors in road traffic introduces new challenges. As the so far unregulated frequency band becomes increasingly crowded, radar sensors suffer from mutual interference between multiple radar sensors. This interference must be mitigated in order to ensure a high and consistent detection sensitivity. In this paper, we propose the use of Complex-Valued Convolutional Neural Networks (CVCNNs) to address the issue of mutual interference between radar sensors. We extend previously developed methods to the complex domain in order to process radar data according to its physical characteristics. This not only increases data efficiency, but also improves the conservation of phase information during filtering, which is crucial for further processing, such as angle estimation. Our experiments show, that the use of CVCNNs increases data efficiency, speeds up network training and substantially improves the conservation of phase information during interference removal.

The critical issue of mutual interference in automotive radar leads to a growing number of processing algorithms for mitigation. Analyses of their performance have been mostly based on either simulations or on a small number of exemplary measurements. In this contribution, a corpus of real-world radar measurements is used, to which strongly varying simulated interferences are added. A statistical analysis framework is then applied to a set of state-of-the-art mitigation algorithms, studying the impact of highly complex object environments on their performance.

Radar sensors are crucial for environment perception of driver assistance systems as well as autonomous cars. Key performance factors are a fine range resolution and the possibility to directly measure velocity. With a rising number of radar sensors and the so far unregulated automotive radar frequency band, mutual interference is inevitable and must be dealt with. Sensors must be capable of detecting, or even mitigating the harmful effects of interference, which include a decreased detection sensitivity. In this paper, we evaluate a Convolutional Neural Network (CNN)-based approach for interference mitigation on real-world radar measurements. We combine real measurements with simulated interference in order to create input-output data suitable for training the model. We analyze the performance to model complexity relation on simulated and measurement data, based on an extensive parameter search. Further, a finite sample size performance comparison shows the effectiveness of the model trained on either simulated or real data as well as for transfer learning. A comparative performance analysis with the state of the art emphasizes the potential of CNN-based models for interference mitigation and denoising of real-world measurements, also considering resource constraints of the hardware., Comment: 2020 IEEE International Radar Conference (RADAR)

The number of cars equipped with radar sensors is steadily increasing. Hence, mutual sensor interference will inevitably occur frequently. Interference results in decreased sensing performance, potentially leading to dangerous traffic situations. For the chirp sequence (CS) radars prevalent in automotive applications, different processing algorithms for interference mitigation have been proposed. However, the performances of these algorithms have not typically been evaluated and compared in a systematic way. In this paper we describe a framework based on the statistical evaluation of performance measures, suitable for a general comparison of mutual radar interference mitigation algorithms. The proposed methodology is then applied to analyze selected algorithms based on radar simulations.

Ray-tracing tools allow for deterministic simulation of the channel impulse response. Studies show that these tools work well when the impulse response consists only of a few distinct components. However, measurements of the channel impulse response in indoor environments show a diffuse tail. This diffuse tail is difficult to include in ray-tracing due to the computational complexity.We propose a hybrid model to include deterministic components and the diffuse tail by combining ray-tracing with a propagation graph. The recursive structure of the propagation graph allows for a computationally efficient calculation of the channel transfer function considering infinitely many components. We use ray-tracing and the theory of room electromagnetics to obtain the parameter settings for the propagation graph. Thus the proposed hybrid model does not require new or additional parameters in comparison to ray-tracing. Simulation results show good agreement with measurements with respect to the inclusion of the diffuse tail in both the delay power spectrum and the azimuth delay power spectrum.

The Cramer–Rao lower bound on the ranging error variance is revisited to quantify the influence of dense multipath in indoor environments. Our analytical results yield novel insight on the scaling of the ranging and positioning accuracy as a function of bandwidth and number of diversity branches. It also yields insight on the detectability of the useful line-of-sight signal component. It is found that the Fisher information scales faster than quadratically in bandwidth but only linearly in the number of independent diversity branches. We investigate the entire bandwidth-range from the flat-fading narrowband case up to ultra-wideband.

Radar sensors are increasingly utilized in today’s cars. This inevitably leads to increased mutual sensor interference and thus a performance decrease, potentially resulting in major safety risks. Understanding signal impairments caused by interference accurately helps to devise signal processing schemes to combat said performance degradation. For the frequency modulated continuous wave (FMCW) radars prevalent in automotive applications, it has been shown that so-called non-coherent interference occurs frequently and results in an increase of the noise floor. In this work we investigate the impact of interference analytically by focusing on its detailed description. We show, among others, that the spectrum of the typical interference signal has a linear phase and a magnitude that is strongly fluctuating with the phase parameters of the time domain interference signal. Analytical results are verified by simulation, highlighting the dependence on the specific phase terms that cause strong deviations from spectral whiteness.

Radar systems are key components for today's advanced driver assistance systems such as adaptive cruise control or emergency brake assistants. Along with the rising utilization of radars in modern cars, the issue of interference amongst themselves arises. It has been shown in previous work that interference between different frequency modulated continuous wave (FMCW) radar systems leads to an increased noise floor. This may severely impact the detectability of objects, especially those with a small radar cross section like pedestrians. In this work we propose a novel concept to mitigate interference in FMCW radar transceivers using digital signal processing. The actual interference cancellation is carried out in frequency domain taking into account a sequence of FMCW chirps. Therewith noise suppression is performed, the interference is cancelled, and the object information is retained in the radar image. In contrast to existing interference cancellation concepts, no threshold needs to be chosen in advance. We prove our method with simulation results, and compare it to existing work.

For making car driving a much safer experience, various forms of radar systems and other object detection systems are being installed. In this paper we consider frequency-modulated continuous-wave (FMCW) automotive radar systems which collect information in so-called Range-Doppler (RD) matrices. The large amount of data collected will soon overwhelm the capacity of automotive data interfaces, and efficient data representation, and compression schemes become increasingly important for the automotive industry. To this end we propose low-rank approximation of RD matrices provided by truncated singular value decomposition. We analytically and numerically investigate error estimates for our approach. We demonstrate that low-rank approximations of RD matrices with a compression factor of over twenty still capture all essential information contained in the original RD matrix.

It is well known that the time-of-flight ranging performance is heavy influenced by multipath propagation within a radio environment. This holds in particular in dense multipath channels as encountered in indoor scenarios. The signal bandwidth has a tremendous influence on this effect, as it determines whether the time resolution is sufficient to resolve the useful line-of-sight (LOS) signal component from interfering multipath. This paper employs a geometry-based stochastic channel model to analyze and characterize the ranging error variance as a function of the bandwidth, covering the narrowband up to the UWB regimes. The Cramer-Rao lower bound (CRLB) is derived for this purpose. It quantifies the impact of bandwidth, SNR, and parameters of the multipath radio channel and can thus be used as an effective and accurate channel model (e.g.) for the cross-layer optimization of positioning systems. Experimental data are analyzed to validate our theoretical results.

We apply joint probabilistic data association (JPDA) to multipath-assisted indoor navigation and tracking (MINT). In MINT, position-related information in multipath components (MPCs) is exploited to increase the accuracy and robustness of indoor tracking. Conventional MINT algorithms are based on deterministic data association and perform a global nearest-neighbor “hard” association of MPC-related delays with the room geometry. In such a setup, incorrect associations may lead to severe tracking errors and to divergence of the Bayesian filter. Here, we propose a JPDA-MINT algorithm that is able to handle difficult situations where MPC delays overlap and data association is ambiguous. The algorithm is based on a recently introduced loopy belief propagation scheme that performs probabilistic data association jointly with agent state estimation, scales well in all relevant systems parameters, and has a very low computational complexity. Using data from an ultra-wideband indoor measurement campaign, we demonstrate that the proposed JPDA-MINT algorithm is highly accurate and more robust than the conventional MINT algorithms based on deterministic data association.

Indoor environments are characterized by harsh multipath conditions. Multipath-assisted indoor navigation and tracking (MINT) exploits position-related features of the propagation channel to improve its accuracy and robustness. In this work we introduce an anchor-free, cooperative MINT algorithm. The algorithm uses monostatic and bistatic (cooperative) measurements conducted by the agents. The estimated multipath components are associated to virtual anchors exploiting their position-related information. We present a proof-of-concept using data from an ultra-wideband measurement campaign, reaching a position accuracy better than 6 cm for 90% of the measurement points.

Highly accurate and reliable indoor positioning — at accuracy levels in the 10 cm range — will enable a large a number of innovative location-based applications because such accuracy levels essentially allow for a useful real-time interaction of humans and cyber-physical systems. Activity recognition, navigation at "shelf" level, geofencing, process monitoring and process control are among the envisioned services that will yield numerous applications in various domains. This paper reviews the difficulties faced by indoor positioning systems, motivating the requirement for a large signal bandwidth and how a lack of bandwidth can be compensated by multi-antenna systems. The potential capabilities of upcoming generations of wireless systems will increasingly make high-accuracy positioning available in near future.

This paper analyzes the achievable ranging and positioning performance for two design constraints in a radio frequency identification (RFID) system: (i) the bandwidth of the transmit signal and (ii) the use of multiple antennas at the readers. The ranging performance is developed for correlated and uncorrelated constituent channels by utilizing a geometry-based stochastic channel model for the downlink and the uplink. The ranging error bound is utilized to compute the precision gain for a ranging scenario with multiple collocated transmit and receive antennas. The position error bound is then split into a monostatic and bistatic component to analyze the positioning performance in a multiple input, multiple output (MIMO) RFID system. Simulation results indicate that the ranging variance is approximately halved when utilizing uncorrelated constituent channels in a monostatic setup. It is shown that both the bandwidth and the number of antennas decrease the error variance roughly quadratically.

The most important factors impairing the performance of radio-based indoor localization systems are propagation effects like strong reflections or diffuse scattering. To the full extent, these effects can be captured only by time-consuming measurement campaigns. Ray tracing (RT) offers the possibility to predict the radio channel for a certain environment, avoiding the need for measurements. However, it is crucial to include all relevant propagation mechanisms in the RT as well as to validate the obtained results. In this paper, we analyze if sub-band divided RT can yield realistic ultra-wideband channel impulse responses. We use the RT results for performance analysis of multipath-assisted localization, which depends directly upon the above mentioned propagation effects. In particular, it has been shown that the ratio of the signal energies of deterministically reflected paths to interfering diffuse components quantifies the amount of position-related information of deterministic multipath components. Comparison of this ratio to measurement data is thus useful to validate the sub-band divided RT. The results highlight the need for proper modeling of the diffuse multipath, as estimates of this energy ratio using RT are often overly optimistic. However, the obtained localization performance predictions using measurements and RT show general agreement.

To obtain good position accuracy with state-of-the-art indoor localization algorithms, multiple anchors must be within radio range of the user. However, anchor placement and maintenance is expensive. By reducing the number of anchors per room, overall costs are reduced. In [1], an algorithm that can estimate the position of a user using only one anchor and a priori floor plan information. However, performance for static localization was poor due to the presence of ambiguities. This paper describes a cooperative positioning algorithm that utilizes a single anchor and provides both the position and the position uncertainty for multiple users, by letting the users exchange their position information. The problem is represented with a factor graph, and belief propagation (BP) is used to extract the positions and their accuracy. The proposed cooperative algorithm leads to a significant improvement of the positioning accuracy compared to the non-cooperative method from [1].