Search Results

With further development of autonomous vehicles additional challenges appear. One of these challenges arises in the context of mixed traffic scenarios where automated and autonomous vehicles coexist with manually operated vehicles as well as other road users such as cyclists and pedestrians. In this evolving landscape, understanding, predicting, and mimicking human driving behavior is becoming not only a challenging but also a compelling facet of autonomous driving research. This is necessary not only for safety reasons, but also to promote trust in artificial intelligence (AI), especially in self-driving cars where trust is often compromised by the opacity of neural network models. The central goal of this study is therefore to address this trust issue. A common approach to imitate human driving behavior through expert demonstrations is imitation learning (IL). However, balancing performance and explainability in these models is a major challenge.

Mechanical systems accomplish their tasks better when enhanced with cyber technologies. With the rapidly escalating desire for high efficiency, optimization and flexibility, these physical systems ought to be integrated with cyber technologies that enhance exhaustive manipulation of resources and productivity. The gateway for such a synergetic integration can be referred to as digitalization. Details regarding the digitization of a High-altitude Simulation chamber are discussed thoroughly in this paper. The simulation chamber was originally designed and developed as a test bench to study the characteristics of alternative fuels used in the engines of handheld tools in different altitudes and thermal conditions. It encompasses all the possible realistic temperature variations with altitude raising to 3500m above sea level.

As automotive systems are becoming increasingly distributed, communication between their components is becoming even more eminent. In safety-relevant distributed systems, the reliability of communication between nodes is crucial for the safety of a system. To guarantee such reliability, it is prerequisite that all nodes in the system have a consistent view of which nodes are functioning correctly and which are not (group membership). In this paper existing algorithms for ensuring group membership are presented and possible solutions for communication systems without such functionality, for example FlexRay, as well as a solution for a network based approach are outlined.

In this paper an approach is presented to determine an adequate number of clusters automatically in case of clustering a distributed automotive electronic system. Hereby, this approach is based on the ISODATA clustering algorithm. Its advantages are its flexibility and less computational effort in comparison to normally used partitioning algorithms. In order to cluster a distributed automotive electronic system with respect to a reduced external communication the input data normally used for partitioning algorithms has to be adapted. Besides, a new overall quality criterion is introduced to validate the results of clustering in reference to the busload before test stage.

Bus systems like CAN or FlexRay allowed great advances in automotive electronics over the last 20 years. In order to function in an environment which requires the communication medium to tolerate one safety-relevant fault, these bus systems require a second, redundant bus to act as a backup for the original unit. With the network approach presented in this paper (SafeNet) it is possible to use the network intrinsic redundancy to keep the network fail-safe after at least one safety relevant fault in the network. To ensure this, messages are relayed to every node in the network. Even though the message delivery times in the network are not deterministic, it is shown that it is suitable for safety-relevant applications like drive-by-wire. Due to the simple point-to-point connections used to connect the nodes, high speeds can be achieved. The network approach is compared to both CAN and FlexRay under different aspects.

In this paper the problem of fault detection in distributed systems is addressed. Due to the trend that these systems are incorporating an increasing number of subsystems from different suppliers fault detection is becoming an essential part of distributed system design. While meeting the typical constraints of the automotive industry there is the demand on increased safety and improved availability. Because of the connection of different subsystems, errors propagate through the system and may affect other subsystems where they can be detected. The key task which is dealt with in this paper is to detect the origin of these errors. Therefore, Hierarchical Colored Bayesian Petri-Nets are introduced to fulfill fault detection according to Bayesian networks. To reduce calculation efforts, the principle of clustering is included.

In this paper an approach to clustering of complex electronic systems using Self-Ordering Maps (SOMs) is presented. SOMs are neural networks which learn through a competitive learning algorithm. In order to use SOMs for the clustering of electronic networks, a representation of the communication behavior in n-dimensional space is developed. The SOM is then used as a nonlinear projection of this space onto a two-dimensional plane. Two examples of clustering are given. The more complex of the two is verified by comparing the behavior of the clustered system and the unclustered system on a simple model of the CAN bus. It is shown that SOMs can be used to effectively cluster complex electronic systems.

There are a number of tools available to assist the engineer during the automotive electronics design process, for example when transferring a graphical specification to a real time rapid prototyping environment. One step in this tool chain however is largely ignored by automated design tools: mapping a large monolithic model to a distributed system, more specifically the mapping of several functions on only a few electronic control units (ECUs) which are connected by a bus. In this paper we will present a method to analyze the underlying functional structure of a given model, partition it using a heuristic algorithm and verify the results with a model of the CAN bus. Based on a given functional model, we will show how to extract an algebraic representation of the communication behavior, the adjacency matrix. Using the adjacency matrix, the heuristic algorithm Best Gain First can be applied to map functions to ECUs.