ML4CPS 2024 : Machine Learning for Cyber-Physical Systems

Cyber-physical systems possess the capability to adjust to evolving demands. When coupled with machine learning, various domains like predictive maintenance, self-optimization, and fault diagnosis spring to mind. An essential condition for realizing this potential is the accessibility of machine learning techniques to engineers.


Therefore, the 7th Machine Learning 4 Cyber Physical Systems - ML4CPS - conference offers researchers and users from various fields an exchange platform. The conference will take place March 2024, 20th till 22th at Fraunhofer Forum in Berlin. Hosts are the Fraunhofer Institute of Optronics, System Technologies and Image Exploitation IOSB and the Helmut Schmidt University (HSU).

Papers may cover, but are not limited to the following topics:
• Large Language Models for CPS: Large language models possess the capability to facilitate human-machine interaction. Their capacity to interpret and generate text unlocks novel opportunities for intelligent automation, contextual comprehension, and seamless integration of various data sources, ultimately enhancing the overall performance and functionality of cyber-physical systems.
• Multimodal Learning: These models possess the capability to facilitate the integration of multiple diverse sensors. Multimodal machine learning models can effectively combine sensor measurements, visual data, contextual information, and various other types of inputs. By using these models, cyber-physical systems can enhance their perception, especially in complex real-world scenarios, leading to improved overall performance.
• Robust Machine Learning: Those techniques play a vital role since they reduce effects of noisy or adversarial data. By incorporating robustness measures, ML models can demonstrate improved resilience, enhanced generalization capabilities, and reliable performance when being used for cyber-physical systems.
• Integrating domain knowledge in neural networks: This is a crucial aspect for building robust and high-performing neural networks. There are several ways to incorporate prior knowledge into the neural network, like designing the network architecture, incorporating additional data from simulations, or imposing constraints on the loss function. All approaches lead to improved network performance and adaptability in cyber-physical systems.

All questions related to paper submissions should be emailed to [email protected].