Utilizing cutting-edge AI to address needs in Health, Robotics, Energy, and Manufacturing
The rapid development of AI (Artificial Intelligence) is continuously influencing various industries, from healthcare to manufacturing. AI-DAPT’s project aims to incorporate MLOps (Machine Learning Operations) and AutoML (Automated Machine Learning) into its framework to innovate, stay in the forefront of AI model implementation and revolutionize the findings across various industries. Our efforts started with examining the most recent scientific and technological innovations in AI, that are critical for the project’s execution and AI-DAPT platform creation. Some of the methods being investigated include data preprocessing, synthetic data generation, explainable AI, first-principles models and hybrid science-guided AI models.
AI-DAPT will be validated through four demonstrators in the fields of Health, Robotics, Energy, and Manufacturing. Using various techniques such as questionnaires, focus groups and workshops, we gathered input from all four demonstrators in an agile way on what they would like the platform to accomplish to meet their needs in their respective fields. We successfully collected these needs (user stories) and analyzed them, assisting us in defining the technical requirements for our platform as the project progresses.
Through intensive investigation on the cutting-edge technologies we identified in our AI research, we aim to identify optimal solutions to effectively meet the needs of our demonstrators. The insights from the four domains highlight the diverse and specific needs that our platform must support. By focusing on these requirements, we can ensure that the AI-DAPT platform can solve real-world problems effectively.
Insights of Demonstrator 1 – Health Domain
The aim of the demonstrator in the health domain is to identify a non-invasive technique to detect and predict type-2 diabetes. Current state-of-the-art diabetes prediction methods, such as measuring fasting blood sugar levels, are intrusive and inconvenient. In contrast, AI-DAPT aims to utilize PPG (photoplethysmography) data, which can be collected easily through the use of smartwatches. Thus, several user stories require the development and validation of such techniques, whilst integrating them into smart watches for ubiquitous diabetes monitoring. Additionally, patients and health professionals expressed the need for monitoring glucose levels under various conditions, such as after eating, while fasting, and during exercise. In addition, privacy of glucose data and adherence to privacy regulations and ethical guidelines are of major concern which we will take into consideration, due to the sensitive nature of medical data.
Insights of Demonstrator 2 – Robotics Domain
Our second demonstrator focuses on robotics and cognitive ergonomics, aiming to optimize working conditions through effective human-machine collaboration. To achieve this, they need to monitor the well-being and stress levels of the robot operators, enabling them to predict errors, enhance productivity and reduce costs. Therefore, the collected requirements involve the utilization of smartwatch data to assess the well-being of an operator in real-time, preparing data to predict mistakes, and monitoring stress levels. Moreover, AI-DAPT plans on automating the pipeline of data collection and prediction. Similarly to the health domain, there is a significant demand for privacy and data anonymization to protect the operators’ information.
Insights of Demonstrator 3 – Energy Domain
Understanding and predicting household energy consumption and costs is essential, especially given the recent surge in energy bills and the inefficiency of the existing EU building stock. Current approaches face challenges, as most solutions like smart thermostats are limited and do not effectively manage demand or consider consumer behavior. Accurate forecasting of energy consumption and peak-load is crucial for optimizing energy use and enhancing efficiency. For this effort, this demonstrator’s user stories emphasized the need for accurate load prediction and energy savings. This involves monitoring and forecasting energy consumption at individual customer endpoints and aggregating the data over larger populations to predict large-scale energy demand.Critical requirements include predicting energy savings, future load prediction, and creating behavioral profiles of energy consumption for dynamic pricing strategies. Ensuring privacy and management of clients’ data is also crucial.
Insights of Demonstrator 4 – Manufacturing Domain
In the manufacturing sector, the focus is on integrating existing software platforms with the AI-DAPT platform and improving maintenance operations. To reduce the operational costs, extend equipment lifespans and enhance operational efficiency, it is important to maintain machines and tools effectively. One of the major elements in the predictive maintenance strategies is accurately predicting the RUL (Remaining Useful Life) of machine components. Therefore, critical requirements include providing status updates on equipment availability, organizing maintenance operations based on scheduling, prioritization and budgeting, and predicting available spare parts for JIT (Just-In-Time) management. There is also a need to identify corrupt data and ensure customer data privacy.
Conclusions
By integrating and analyzing the identified pilot and user needs, the AI-DAPT project aims to develop a platform that not only meets technical requirements but also addresses the diverse challenges of each field. Our primary focus is to leverage state-of-the-art AI technologies to tackle these challenges. This collaborative approach ensures that the AI-DAPT platform will be robust, user-centric, and capable of driving innovation across multiple industries.
Stay tuned for more updates as we progress further in the AI-DAPT project!