Michael Ulrich

• University of Stuttgart

Masked autoencoders (MAE) have shown tremendous potential for self-supervised learning (SSL) in vision and beyond. However, point clouds from LiDARs used in automated driving are particularly challenging for MAEs since large areas of the 3D volume are empty. Consequently, existing work suffers from leaking occupancy information into the decoder and...

This paper introduces the Bosch street dataset (BSD), a novel multi-modal large-scale dataset aimed at promoting highly automated driving (HAD) and advanced driver-assistance systems (ADAS) research. Unlike existing datasets, BSD offers a unique integration of high-resolution imaging radar, lidar, and camera sensors, providing unprecedented 360-deg...

Having precise perception of the environment is crucial for ensuring the secure and reliable functioning of autonomous driving systems. Radar object detection networks are one fundamental part of such systems. CNN-based object detectors showed good performance in this context, but they require large compute resources. This paper investigates sparse...

We propose DeepFusion, a modular multi-modal architecture to fuse lidars, cameras and radars in different combinations for 3D object detection. Specialized feature extractors take advantage of each modality and can be exchanged easily, making the approach simple and flexible. Extracted features are transformed into bird's-eye-view as a common repre...

This paper presents a method to learn the Cartesian velocity of objects using an object detection network on automotive radar data. The proposed method is self-supervised in terms of generating its own training signal for the velocities. Labels are only required for single-frame, oriented bounding boxes (OBBs). Labels for the Cartesian velocities o...

This paper presents novel hybrid architectures that combine grid- and point-based processing to improve the detection performance and orientation estimation of radar-based object detection networks. Purely grid-based detection models operate on a bird's-eye-view (BEV) projection of the input point cloud. These approaches suffer from a loss of detai...

This paper presents an novel object type classification method for automotive applications which uses deep learning with radar reflections. The method provides object class information such as pedestrian, cyclist, car, or non-obstacle. The method is both powerful and efficient, by using a light-weight deep learning approach on reflection level rada...

This article investigates the parameter coupling of range, Doppler and direction-of-arrival (DOA) for Wavelength Diverse (WLD) MIMO radar with time-division multiplexing (TDM). In particular, the Cramer-Rao Bound for such a WLDMIMO system is calculated and used for a novel array and wavelength design approach. Furthermore, a multi-target DOA estima...

Multi-carrier (MC) multiple-input multiple-output (MIMO) radar offers an additional degree of freedom in the array optimization through the carrier frequencies. In this paper, we study the MC-MIMO array optimization with respect to the direction of arrival (DOA) estimation based on the Cramer-Rao bound (CRB). In particular, we choose the transmit a...

We present in this work a novel approach for the reconstruction of wired network topologies from reflection measurements. Existing approaches state the network reconstruction as discrete optimization problem, which is difficult to solve. The (discrete) topology is optimized while the cable lengths are a secondary result. The contribution of this p...

Amplitude and phase imbalance (IQ imbalance) seriously degrades the performance of FMCW radar systems with quadrature demodulator. The non-ideality of the radar signals has two main effects: The single-channel IQ imbalance leads to an image peak in the FMCW spectrum, resulting in wrong detections. The IQ imbalance between channels affects the direc...

Simulation is a powerful tool for radar modulation and array design, the development of radar signal processing algorithms and the evaluation of defined scenarios. With the evolution of radar hardware, the range resolution of radar is often much smaller than the size of the radar targets which cannot be treated as point targets any more. In the lit...

Multi-carrier (MC) multiple-input multiple-output (MIMO) radar was recently applied to build sparse virtual arrays with a large aperture for a high-accuracy direction-of-arrival (DOA) estimation. The resulting grating lobes (DOA ambiguities) were resolved using multiple carriers. One problem of MC-MIMO is the coupling of the unknown parameters rang...

In this paper we propose a multi-carrier (MC) based sparse array for improved direction-of-arrival (DOA) estimation. We use a spatial (physical or virtual) array of a small number of antennas, but with a large aperture to achieve a high DOA estimation accuracy. The resulting problem of grating lobes (spatial aliasing) is addressed by using multiple...

We present in this paper a novel algorithm CoMaTeCh for the inference of wired network topology using reflection measurements at multiple cable ends. This is useful for applications where the topology of an existing wired network (e.g. communication networks, powerline networks) is unknown and needs to be reconstructed in a non-intrusive way. Start...