Logo: to the web site of Uppsala University

The integration of artificial intelligence into the development and production of mechatronic products offers a substantial opportunity to enhance efficiency, adaptability, and system performance. This paper examines the utilization of reinforcement learning as a control strategy, with a particular focus on its deployment in pivotal stages of the product development lifecycle, specifically between system architecture and system integration and verification. A controller based on reinforcement learning was developed and evaluated in comparison to traditional proportional-integral controllers in dynamic and fault-prone environments. The results illustrate the superior adaptability, stability, and optimization potential of the reinforcement learning approach, particularly in addressing dynamic disturbances and ensuring robust performance. The study illustrates how reinforcement learning can facilitate the transition from conceptual design to implementation by automating optimization processes, enabling interface automation, and enhancing system-level testing. Based on the aforementioned findings, this paper presents future directions for research, which include the integration of domain-specific knowledge into the reinforcement learning process and the validation of this process in real-world environments. The results underscore the potential of artificial intelligence-driven methodologies to revolutionize the design and deployment of intelligent mechatronic systems.

Mechatronic product development is a complex and multidisciplinary field that encompasses various domains, including, among others, mechanical engineering, electrical engineering, control theory and software engineering. The integration of artificial intelligence technologies is revolutionizing this domain, offering opportunities to enhance design processes, optimize performance, and leverage vast amounts of knowledge. However, human expertise remains essential in contextualizing information, considering trade-offs, and ensuring ethical and societal implications are taken into account. This paper therefore explores the existing literature regarding the application of artificial intelligence as a comprehensive database, decision support system, and modeling tool in mechatronic product development. It analyzes the benefits of artificial intelligence in enabling domain linking, replacing human expert knowledge, improving prediction quality, and enhancing intelligent control systems. For this purpose, a consideration of the V-cycle takes place, a standard in mechatronic product development. Along this, an initial assessment of the AI potential is shown and important categories of AI support are formed. This is followed by an examination of the literature with regard to these aspects. As a result, the integration of artificial intelligence in mechatronic product development opens new possibilities and transforms the way innovative mechatronic systems are conceived, designed, and deployed. However, the approaches are only taking place selectively, and a holistic view of the development processes and the potential for robust and context-sensitive artificial intelligence along them is still needed.

Steer-by-wire systems represent a key technology for highly automated and autonomous driving. In this context, robust steering control is a fundamental precondition for automated vehicle lateral control. However, there is a need for improvement due to degrees of freedom, signal delays, and nonlinear characteristics of the plant which are unconsidered in the design models for the design of current steering controls. To be able to design an extremely robust steering control, suitable optimal models of a steer-by-wire system are required. Therefore, this paper presents an innovative nonlinear detail model of a steer-by-wire system. The detail model represents all characteristics of a real steer-by-wire system. In the context of a dominance analysis of the detail model, all dominant characteristics of a steer-by-wire system, including parameter dependencies, are identified. Through model reduction, a reduced model of the steer-by-wire system is then developed that can be used for a subsequent robust control design. Furthermore, this paper compares the steer-by-wire system with a conventional electromechanical power steering and shows similarities as well as differences.

Advanced driver assistance systems are an important step on the way towards the autonomous driving. However, there are new challenges in the release of increasingly complex systems. For the testing of those systems many test kilometers are necessary to represent sufficient diversity. Hence, the virtual testing of driver assistance systems brings new opportunities. In virtual environments, it is possible to run a much higher distance in a short time. Simultaneously, the complexity of the environment and the test scenarios are individually adjustable. It is possible to test scenarios that are not feasible in a real environment due to a risk of injury. A big challenge is the physical correct implementation of real vehicles and their components into the Virtual Reality. To enable a realistic virtual testing the vehicles surrounding sensors need to be modeled adequately. Thus, this paper presents an approach for the implementation of a Lidar model into a Virtual Reality. A physical Lidar model is combined with a real-time capable vehicle dynamics model to investigate the influence of vehicle movements to the sensor measurements. The models are implemented into a highly realistic virtual city environment. Finally, a test campaign shows the influence of the Lidars physics and the vehicle dynamics on the detection results.

On the way to highly automated and autonomous driving, a robustly designed steering system is a key component. Therefore, this article presents a direct discrete control design for modern steer-by-wire systems. The novel approach consists of a true multivariable control for both the driver´s steering torque and the rack position simultaneously using the requested torques of the downstream (AU) and upstream (FU) motor as control variables. For the control design, an optimal reduced plant model is used. It is derived from a detailed model of a steer-by-wire system with nine degrees of freedom. The reduced plant model is augmented by linear models for the reference and disturbance environment of the steer-by-wire system as well as discretized based on the characteristics of the input variables. For this augmented model, a direct discrete multivariable linear quadratic Gaussian (LQG) compensator design is performed. The proposed control design considers the entire environment of the real steering system. The direct discrete approach restores the good characteristics of the continuous control and ensures that the discretization does not have any adverse effects. As a result, the resulting discrete control system shows the same good dynamic characteristics as the continuous system and has excellent robustness characteristics. Hence, the presented control satisfies the requirements of a modern steering system and can be adapted to various driving situations.

In the fast-paced world of automotive development, the need for effective information exchange between different domains is becoming increasingly critical. The parallelized and agile nature of development demands early and frequent access to accurate information across various departments. However, deviations and uncertainties in data can lead to costly disruptions in the development process. Artificial intelligence based prediction models offer a potential solution by providing estimates based on previous projects or analyzed products, even in the early stages of development. While rough estimates may suffice initially, it is important to understand the accuracy of such predictions. This study therefore aims to evaluate the performance characteristics of different uncertainty analysis methods and assess their applicability in agile automotive development processes. By considering the specific requirements and constraints of each method, a decision tree is proposed to recommend suitable and situation-appropriate methods for performing uncertainty analyses in network prediction. The goal is to enhance data exchange between departments, mitigate disruptions, and ensure informed decision-making throughout the development process.