Heinrich GOTZIG - Academia.edu (original) (raw)

Papers by Heinrich GOTZIG

In this paper we present a novel machine learning approach for noise suppression in the signal ge... more In this paper we present a novel machine learning approach for noise suppression in the signal generated by automotive industrial grade active ultrasonic sensors. A convolutional neural network (CNN) based machine learning approach is presented. State of art noise suppression methods are also discussed and used as benchmark against the proposed machine learning approach. The results of numerous simulated scenarios as well as actual sensor measurement campaigns are presented and discussed. Several metrics are derived to quantify the quality of the signal and give an indication of the performance of the different approaches of noise suppression. These derived metrics assess the performance of the different approaches in terms of amount of noise suppressed and amount of distortion introduced to the signal of interest.

This paper provides an experimental performance analysis of advanced layers for electronic circui... more This paper provides an experimental performance analysis of advanced layers for electronic circuit boards. The base material of the advanced layers used in the experiments is Al-5052, with the dimensions of 100mm x 20mm x 1mm. As insulating material between the base and conductor, two aluminium coatings were investigated and the surface structure and thermal conductivity of the coatings were examined. Furthermore, three electroconductive adhesives were applied onto the coatings. Initially, electrical performance is investigated and electrical components were applied onto the new surface structure. Finally, the environmental tests were performed to determine the robustness of the evaluated system.

This paper proposes the replacement of a conventional Insulating-Metal-Substrate (IMS) dielectric... more This paper proposes the replacement of a conventional Insulating-Metal-Substrate (IMS) dielectric layer with an electroconductive adhesive which is applied onto a treated aluminium heat sink. The adhesive based structure results in a lower thermal resistance of the system compared to IMS based on Aluminium. The model covers the investigation of two IMS structure with a Copper conductor layer with a thickness of 200µm, a dielectric layer of 50µm with different thermal conductivities and an Aluminium base layer with a thickness of 920µm. However, the IMS is placed onto a 100 x 100 x 33 mm Aluminium heat sink. The investigated adhesive based structure consists of the same Aluminium heat sink treated with an oxidation layer with a thickness of 50µm and an adhesive conductor layer with 30µm thickness. Furthermore, the use of forced convection to reduce the junction temperature of all three models is also considered.

Springer eBooks, Dec 8, 2015

In this paper we present a novel approach for using industrial grade ultrasonic sensors to perfor... more In this paper we present a novel approach for using industrial grade ultrasonic sensors to perform echolocation by detecting ultrasonic echoes using machine learning. We show how this methodology is robust against the influence of ultrasonic noise sources in the environment as well as undesired reflections coming from the terrain. Several noise sources and noise powers are assessed as well as different terrain types. The results are benchmarked against the state of art energy thresholding and matched filters correlation algorithms clearly showing the superiority of the machine learning based approach.

This paper presents an efficient model for combining automotive trajectory planning with predicte... more This paper presents an efficient model for combining automotive trajectory planning with predicted environment interactions, named progressively interacting trajectories (PITRA). The model allows to plan trajectories for fullyautomated vehicles by actively considering how other traffic participants will react to the trajectory, while retaining many of the advantages of variational trajectory optimization methods, in particular expressiveness and ease of computation. This enables maneuvers such as proactively claiming a gap during a lane change in dense traffic, which are impossible to model in classical variational models. The PITRA approach does not rely on a specific prediction model, but can be used in combination with a wide range of existing models. Its model assumptions and limitations are derived theoretically and demonstrated in several realistic scenarios. * ξ minimizes S. A key advantage of variational methods is the way they can model desired and undesired properties of trajectories in a unified way, which is intuitive, expressive, computationally tractable and rather well-understood due to its long-established physical relevance. These models have been advanced in recent publications, to replace purely local trajectory optimization by approximate global optimization (cf. [3], [4]), search space heuristics (cf. [5]) and fail-safe emergency trajectories (cf. [6]). However, a feature that is both in considerable demand (voiced e.g. in the peer reviews to our previous publication, [6]) and still absent in the described variational models, is the ability to consider interaction during trajectory planning, in the sense that explored trajectory candidates are evaluated with respect to the reactions they cause in other traffic *This work has been jointly supported by the State of Baden-Württemberg and the Fraunhofer-Gesellschaft under Grant 7-4332-62-FhG/26.

Springer eBooks, Dec 8, 2015

This paper names several risks with iterative optimization of vehicle trajectories modeled throug... more This paper names several risks with iterative optimization of vehicle trajectories modeled through the calculus of variations, as used recently in several approaches to automated driving, and demonstrates how these can be addressed by global optimization, while retaining the original, variational optimization goals. A method for transforming variational methods into Hidden Markov Models is applied to multilane scenarios, such as highways or larger city streets. A particular side-effect of this method is the capability of providing a fail-safe emergency trajectory at almost no further computational effort. The proposed global optimization is demonstrated on several real-world situations and compared to a ground truth of human maneuvers.

Fibers, May 17, 2016

The hygrothermal aging of short glass fiber-reinforced polyamide 6 materials (PA6 GF) represents ... more The hygrothermal aging of short glass fiber-reinforced polyamide 6 materials (PA6 GF) represents a major problem, especially in thin-walled components, such as in the automotive sector. In this study, therefore, the thickness and the glass fiber content of PA6 GF materials were varied and the materials were exposed to hygrothermal aging. The temperature and relative humidity were selected in the range from´40˝C up to 85˝C, and from 10% up to 85% relative humidity (RH). In the dry-as-molded state, the determined Poisson's ratio of the PA6 GF materials was correlated with the fiber orientation based on computer tomography (MicroCT) data and shows a linear dependence with respect to the fiber orientation along and transverse to the flow direction of the injection molding process. With hygrothermal aging, the value of Poisson's ratio increases in the flow direction in the same way as it decreases perpendicular to the flow direction due to water absorption.

This paper provides a steady-state thermal characterisation of advanced insulating layers for iso... more This paper provides a steady-state thermal characterisation of advanced insulating layers for isotropic electroconductive adhesive based Metal Core Printed Circuit Boards. Thermal measurements were conducted using the Thermal Transient Tester to analyse the properties of layers applied onto two different base materials-aluminium and copper. For each base material, four types of insulating layers were characterised. Size of the specimen plate for both base materials was selected as 100mm x 20mm x 1mm. The isotropic electroconductive adhesive was applied onto insulating layers to ensure electrical connection between seven SMD heat-generating components. These components are mounted in the identical distance to each other. The thermal characterisation of power SMD components and the assembling process based on isotropic electroconductive adhesive were evaluated. Application of advanced insulating layer was analysed to assess the thermal performance of complete MCPCB structure. It has been shown that the lowest thermal resistance of the aluminium based insulating layers is 21.1K/W, whereas the lowest thermal resistance of the copper based insulating layer is 15.5K/W.

This paper establishes a duality between the calculus of variations, an increasingly common metho... more This paper establishes a duality between the calculus of variations, an increasingly common method for trajectory planning, and Hidden Markov Models (HMMs), a common probabilistic graphical model with applications in artificial intelligence and machine learning. This duality allows findings from each field to be applied to the other, namely providing an efficient and robust global optimization tool and machine learning algorithms for variational problems, and fast local solution methods for large state-space HMMs.

arXiv (Cornell University), Jul 17, 2023

LiDAR is crucial for robust 3D scene perception in autonomous driving. LiDAR perception has the l... more LiDAR is crucial for robust 3D scene perception in autonomous driving. LiDAR perception has the largest body of literature after camera perception. However, multi-task learning across tasks like detection, segmentation, and motion estimation using LiDAR remains relatively unexplored, especially on automotive-grade embedded platforms. We present a real-time multi-task convolutional neural network for LiDAR-based object detection, semantics, and motion segmentation. The unified architecture comprises a shared encoder and task-specific decoders, enabling joint representation learning. We propose a novel Semantic Weighting and Guidance (SWAG) module to transfer semantic features for improved object detection selectively. Our heterogeneous training scheme combines diverse datasets and exploits complementary cues between tasks. The work provides the first embedded implementation unifying these key perception tasks from LiDAR point clouds achieving 3ms latency on the embedded NVIDIA Xavier platform. We achieve state-of-the-art results for two tasks, semantic and motion segmentation, and close to state-of-the-art performance for 3D object detection. By maximizing hardware efficiency and leveraging multi-task synergies, our method delivers an accurate and efficient solution tailored for real-world automated driving deployment. Qualitative results can be seen at https://youtu.be/H-hWRzv2lIY.

Atz - Automobiltechnische Zeitschrift, Dec 1, 2009

An der Grundfunktionalität von Parkhilfesystemen-der Information des Fahrers über Abstände naher ... more An der Grundfunktionalität von Parkhilfesystemen-der Information des Fahrers über Abstände naher Objekte während des Einparkvorganges-hat sich seit der ersten Generation nur wenig geändert. Valeo war Vorreiter auf diesem Gebiet und hat Anfang der 1990er-Jahre das erste auf Ultraschall basierende Parkassistenzsystem im BMW 7er eingeführt. Im Folgenden werden Entwicklungsstufen, Marktstudien, Wünsche der Endkunden an künftige Parksysteme und deren technologische Umsetzung beschrieben. E n T w I c k l u n g ATZ 12I2009 Jahrgang 111 932 Lenkung Download des Beitrags unter www.ATZonline.de Read the English e-magazine.

This paper presents and evaluates a differential geometric model for single and double track vehi... more This paper presents and evaluates a differential geometric model for single and double track vehicles, which allows to estimate internal state parameters (steering angles, velocities, wheel speeds, ...) from a given planar trajectory over time. Possible applications include accident reconstruction from video footage, constraining trajectory optimization by physical limits, and extracting control parameters a-priori from trajectories, for use in open-loop (non-feedback) controllers for short intervals. The model is easy to compute, yet the evaluation with IPG CarMaker suggests a good level of accuracy.

2023 IEEE 97th Vehicular Technology Conference (VTC2023-Spring)

2022 8th International Conference on Event-Based Control, Communication, and Signal Processing (EBCCSP)

Spiking Neural Networks are a recent and new neural network design approach that promises tremend... more Spiking Neural Networks are a recent and new neural network design approach that promises tremendous improvements in power efficiency, computation efficiency, and processing latency. They do so by using asynchronous spikebased data flow, event-based signal generation, processing, and modifying the neuron model to resemble biological neurons closely. While some initial works have shown significant initial evidence of applicability to common deep learning tasks, their applications in complex real-world tasks has been relatively low. In this work, we first illustrate the applicability of spiking neural networks to a complex deep learning task namely Lidar based 3D object detection for automated driving. Secondly, we make a step-by-step demonstration of simulating spiking behavior using a pre-trained convolutional neural network. We closely model essential aspects of spiking neural networks in simulation and achieve equivalent run-time and accuracy on a GPU. When the model is realized on a neuromorphic hardware, we expect to have significantly improved power efficiency.

In this paper we present a novel machine learning approach for noise suppression in the signal ge... more In this paper we present a novel machine learning approach for noise suppression in the signal generated by automotive industrial grade active ultrasonic sensors. A convolutional neural network (CNN) based machine learning approach is presented. State of art noise suppression methods are also discussed and used as benchmark against the proposed machine learning approach. The results of numerous simulated scenarios as well as actual sensor measurement campaigns are presented and discussed. Several metrics are derived to quantify the quality of the signal and give an indication of the performance of the different approaches of noise suppression. These derived metrics assess the performance of the different approaches in terms of amount of noise suppressed and amount of distortion introduced to the signal of interest.

This paper provides an experimental performance analysis of advanced layers for electronic circui... more This paper provides an experimental performance analysis of advanced layers for electronic circuit boards. The base material of the advanced layers used in the experiments is Al-5052, with the dimensions of 100mm x 20mm x 1mm. As insulating material between the base and conductor, two aluminium coatings were investigated and the surface structure and thermal conductivity of the coatings were examined. Furthermore, three electroconductive adhesives were applied onto the coatings. Initially, electrical performance is investigated and electrical components were applied onto the new surface structure. Finally, the environmental tests were performed to determine the robustness of the evaluated system.

This paper proposes the replacement of a conventional Insulating-Metal-Substrate (IMS) dielectric... more This paper proposes the replacement of a conventional Insulating-Metal-Substrate (IMS) dielectric layer with an electroconductive adhesive which is applied onto a treated aluminium heat sink. The adhesive based structure results in a lower thermal resistance of the system compared to IMS based on Aluminium. The model covers the investigation of two IMS structure with a Copper conductor layer with a thickness of 200µm, a dielectric layer of 50µm with different thermal conductivities and an Aluminium base layer with a thickness of 920µm. However, the IMS is placed onto a 100 x 100 x 33 mm Aluminium heat sink. The investigated adhesive based structure consists of the same Aluminium heat sink treated with an oxidation layer with a thickness of 50µm and an adhesive conductor layer with 30µm thickness. Furthermore, the use of forced convection to reduce the junction temperature of all three models is also considered.

Springer eBooks, Dec 8, 2015

In this paper we present a novel approach for using industrial grade ultrasonic sensors to perfor... more In this paper we present a novel approach for using industrial grade ultrasonic sensors to perform echolocation by detecting ultrasonic echoes using machine learning. We show how this methodology is robust against the influence of ultrasonic noise sources in the environment as well as undesired reflections coming from the terrain. Several noise sources and noise powers are assessed as well as different terrain types. The results are benchmarked against the state of art energy thresholding and matched filters correlation algorithms clearly showing the superiority of the machine learning based approach.

This paper presents an efficient model for combining automotive trajectory planning with predicte... more This paper presents an efficient model for combining automotive trajectory planning with predicted environment interactions, named progressively interacting trajectories (PITRA). The model allows to plan trajectories for fullyautomated vehicles by actively considering how other traffic participants will react to the trajectory, while retaining many of the advantages of variational trajectory optimization methods, in particular expressiveness and ease of computation. This enables maneuvers such as proactively claiming a gap during a lane change in dense traffic, which are impossible to model in classical variational models. The PITRA approach does not rely on a specific prediction model, but can be used in combination with a wide range of existing models. Its model assumptions and limitations are derived theoretically and demonstrated in several realistic scenarios. * ξ minimizes S. A key advantage of variational methods is the way they can model desired and undesired properties of trajectories in a unified way, which is intuitive, expressive, computationally tractable and rather well-understood due to its long-established physical relevance. These models have been advanced in recent publications, to replace purely local trajectory optimization by approximate global optimization (cf. [3], [4]), search space heuristics (cf. [5]) and fail-safe emergency trajectories (cf. [6]). However, a feature that is both in considerable demand (voiced e.g. in the peer reviews to our previous publication, [6]) and still absent in the described variational models, is the ability to consider interaction during trajectory planning, in the sense that explored trajectory candidates are evaluated with respect to the reactions they cause in other traffic *This work has been jointly supported by the State of Baden-Württemberg and the Fraunhofer-Gesellschaft under Grant 7-4332-62-FhG/26.

Springer eBooks, Dec 8, 2015

This paper names several risks with iterative optimization of vehicle trajectories modeled throug... more This paper names several risks with iterative optimization of vehicle trajectories modeled through the calculus of variations, as used recently in several approaches to automated driving, and demonstrates how these can be addressed by global optimization, while retaining the original, variational optimization goals. A method for transforming variational methods into Hidden Markov Models is applied to multilane scenarios, such as highways or larger city streets. A particular side-effect of this method is the capability of providing a fail-safe emergency trajectory at almost no further computational effort. The proposed global optimization is demonstrated on several real-world situations and compared to a ground truth of human maneuvers.

Fibers, May 17, 2016

The hygrothermal aging of short glass fiber-reinforced polyamide 6 materials (PA6 GF) represents ... more The hygrothermal aging of short glass fiber-reinforced polyamide 6 materials (PA6 GF) represents a major problem, especially in thin-walled components, such as in the automotive sector. In this study, therefore, the thickness and the glass fiber content of PA6 GF materials were varied and the materials were exposed to hygrothermal aging. The temperature and relative humidity were selected in the range from´40˝C up to 85˝C, and from 10% up to 85% relative humidity (RH). In the dry-as-molded state, the determined Poisson's ratio of the PA6 GF materials was correlated with the fiber orientation based on computer tomography (MicroCT) data and shows a linear dependence with respect to the fiber orientation along and transverse to the flow direction of the injection molding process. With hygrothermal aging, the value of Poisson's ratio increases in the flow direction in the same way as it decreases perpendicular to the flow direction due to water absorption.

This paper provides a steady-state thermal characterisation of advanced insulating layers for iso... more This paper provides a steady-state thermal characterisation of advanced insulating layers for isotropic electroconductive adhesive based Metal Core Printed Circuit Boards. Thermal measurements were conducted using the Thermal Transient Tester to analyse the properties of layers applied onto two different base materials-aluminium and copper. For each base material, four types of insulating layers were characterised. Size of the specimen plate for both base materials was selected as 100mm x 20mm x 1mm. The isotropic electroconductive adhesive was applied onto insulating layers to ensure electrical connection between seven SMD heat-generating components. These components are mounted in the identical distance to each other. The thermal characterisation of power SMD components and the assembling process based on isotropic electroconductive adhesive were evaluated. Application of advanced insulating layer was analysed to assess the thermal performance of complete MCPCB structure. It has been shown that the lowest thermal resistance of the aluminium based insulating layers is 21.1K/W, whereas the lowest thermal resistance of the copper based insulating layer is 15.5K/W.

This paper establishes a duality between the calculus of variations, an increasingly common metho... more This paper establishes a duality between the calculus of variations, an increasingly common method for trajectory planning, and Hidden Markov Models (HMMs), a common probabilistic graphical model with applications in artificial intelligence and machine learning. This duality allows findings from each field to be applied to the other, namely providing an efficient and robust global optimization tool and machine learning algorithms for variational problems, and fast local solution methods for large state-space HMMs.

arXiv (Cornell University), Jul 17, 2023

LiDAR is crucial for robust 3D scene perception in autonomous driving. LiDAR perception has the l... more LiDAR is crucial for robust 3D scene perception in autonomous driving. LiDAR perception has the largest body of literature after camera perception. However, multi-task learning across tasks like detection, segmentation, and motion estimation using LiDAR remains relatively unexplored, especially on automotive-grade embedded platforms. We present a real-time multi-task convolutional neural network for LiDAR-based object detection, semantics, and motion segmentation. The unified architecture comprises a shared encoder and task-specific decoders, enabling joint representation learning. We propose a novel Semantic Weighting and Guidance (SWAG) module to transfer semantic features for improved object detection selectively. Our heterogeneous training scheme combines diverse datasets and exploits complementary cues between tasks. The work provides the first embedded implementation unifying these key perception tasks from LiDAR point clouds achieving 3ms latency on the embedded NVIDIA Xavier platform. We achieve state-of-the-art results for two tasks, semantic and motion segmentation, and close to state-of-the-art performance for 3D object detection. By maximizing hardware efficiency and leveraging multi-task synergies, our method delivers an accurate and efficient solution tailored for real-world automated driving deployment. Qualitative results can be seen at https://youtu.be/H-hWRzv2lIY.

Atz - Automobiltechnische Zeitschrift, Dec 1, 2009

An der Grundfunktionalität von Parkhilfesystemen-der Information des Fahrers über Abstände naher ... more An der Grundfunktionalität von Parkhilfesystemen-der Information des Fahrers über Abstände naher Objekte während des Einparkvorganges-hat sich seit der ersten Generation nur wenig geändert. Valeo war Vorreiter auf diesem Gebiet und hat Anfang der 1990er-Jahre das erste auf Ultraschall basierende Parkassistenzsystem im BMW 7er eingeführt. Im Folgenden werden Entwicklungsstufen, Marktstudien, Wünsche der Endkunden an künftige Parksysteme und deren technologische Umsetzung beschrieben. E n T w I c k l u n g ATZ 12I2009 Jahrgang 111 932 Lenkung Download des Beitrags unter www.ATZonline.de Read the English e-magazine.

This paper presents and evaluates a differential geometric model for single and double track vehi... more This paper presents and evaluates a differential geometric model for single and double track vehicles, which allows to estimate internal state parameters (steering angles, velocities, wheel speeds, ...) from a given planar trajectory over time. Possible applications include accident reconstruction from video footage, constraining trajectory optimization by physical limits, and extracting control parameters a-priori from trajectories, for use in open-loop (non-feedback) controllers for short intervals. The model is easy to compute, yet the evaluation with IPG CarMaker suggests a good level of accuracy.

2023 IEEE 97th Vehicular Technology Conference (VTC2023-Spring)

2022 8th International Conference on Event-Based Control, Communication, and Signal Processing (EBCCSP)

Spiking Neural Networks are a recent and new neural network design approach that promises tremend... more Spiking Neural Networks are a recent and new neural network design approach that promises tremendous improvements in power efficiency, computation efficiency, and processing latency. They do so by using asynchronous spikebased data flow, event-based signal generation, processing, and modifying the neuron model to resemble biological neurons closely. While some initial works have shown significant initial evidence of applicability to common deep learning tasks, their applications in complex real-world tasks has been relatively low. In this work, we first illustrate the applicability of spiking neural networks to a complex deep learning task namely Lidar based 3D object detection for automated driving. Secondly, we make a step-by-step demonstration of simulating spiking behavior using a pre-trained convolutional neural network. We closely model essential aspects of spiking neural networks in simulation and achieve equivalent run-time and accuracy on a GPU. When the model is realized on a neuromorphic hardware, we expect to have significantly improved power efficiency.