loading the future of Autonomous Mobile Robots


Reach out and we will get back to you!

    Predictive Analytics for AMRs: Insights from the AI REGIO Experiment

    The integration of artificial intelligence (AI) into manufacturing processes has revolutionised how industries operate. The AI REGIO project, funded under the European Union’s H2020 Research and Innovation Programme, has been instrumental in fostering collaboration among Digital Innovation Hubs (DIHs) across Europe. In partnership with arculus, they launched an experiment to optimise robotic systems through predictive analytics and maintenance. This blog post explores the objectives and outcomes of this groundbreaking experience conducted by arculus and AI Regio in cooperation with Forschungszentrum Informatik (FZI).

    What is Predictive Analytics?

    Predictive analytics uses statistics, Artificial Intelligence (AI), data mining, and modelling techniques to forecast future outcomes. This method analyses current and historical data patterns to determine if similar events are likely to happen in the future. This allows businesses to adjust their strategy, to take advantage and prepare for probable future occurrences, leading to improved operational efficiency and reduced risk.

    How does Predictive Analytics work?

    Data scientists utilise predictive models to find connections between various components within specific datasets. Following the collection of data, a statistical model is created, trained and adjusted, in order to generate predictions.

    Predictive Maintenance

    Predictive models have a wide range of applications across numerous and varied fields. Primary use cases include weather forecasting, customer service, voice-to-text translation, and investment portfolio strategies. In the scope of the experiment conducted by AI REGIO with arculus and FZI, the focus was on the so-called predictive maintenance.

    In the background, there is a blurred image of a man in a suit pointing his finger at visual representations of the elements that constitute predictive analytics. Namely: Pattern recognition, artificial intelligence, automation, neural networks, algorithm, data mining, and problem solving.

    Predictive maintenance uses advanced algorithms to anticipate and prevent equipment malfunctions. By analysing patterns and trends in historical data, predictive maintenance models can forecast when machines are likely to experience a breakdown. This data-driven approach enables timely interventions, optimises maintenance schedules, minimises unplanned downtime, and reduces overall repair costs, by addressing issues before they escalate.

    A report by the Deloitte Analytics Institute quantifies the benefits:

    “On average, predictive maintenance increases productivity by 25%, reduces breakdowns by 70% and lowers maintenance costs by 25%. It is based on advanced analytics and marks a new way of organising and implementing maintenance on an industrial scale.”

    The Experiment

    The main goal of the experiment conducted by AI REGIO with arculus, was to leverage predictive maintenance models to make robots more sturdy and efficient. In this context, instead of focusing on optimising control algorithms typically associated with robot navigation, the aim was to improve and optimise processes that could significantly impact arculus’ scaling efforts. These include, among others:

    • Optimisation of maintenance intervals;
    • Reduction of regular service intervals (like scanner cleaning) by making service calls on a need basis rather than schedule-basis;
    • Early detection of sensor faults (Lidar, IMU, Camera) to ensure proactive resolution of issues.
    Close up of the arculee showing its Lidar sensors.
    Proper functioning of the arculee’s Lidar sensors is essential for navigation

    The following requirements were defined as success metrics:

    • Accurate and securely repeatable predictions of specific situations;
    • Continuous and automatable data pipeline from customer projects.

    The results

    The experiment yielded promising results. Firstly, the team successfully connected Internet of Things (IoT) data streams from the robots by utilising a Machine Learning (ML) technique known as automatic schema recognition. This allowed for seamless integration and analysis of the data collected. Furthermore, by creating processing pipelines built on both rule-based and learning-based algorithms, the experiment enabled efficient analysis and prediction of maintenance needs.

    The study also established a model repository, featuring pre-trained models designed to facilitate predictive analytics for Autonomous Mobile Robots (AMRs). These models trained using ML techniques on extensive historical data and provided accurate predictions of maintenance requirements. Lastly, the experiment introduced a pipeline element generator to optimise the maintenance process further. This tool allows for the quick creation of new models tailored to specific use cases through a technique called few-shot learning.

    Close up of an engineer's hands using a machine to gather electric waves from the arculee's control unit.
    Data streams from the robot are a key part of predictive modelling

    Next Steps

    The next step in the experiment will involve developing a high-performance algorithm based on real-life data. This algorithm will focus on detecting dirt on the laser scanners of the AMRs, enabling proactive notification to the maintenance team for prompt cleaning. With this approach, the trial aims to ensure the AMRs’ optimal performance and longevity, as well as minimise any potential operational disruptions they may face.

    Meet the Players

    AI Regio

    The goal of AI REGIO is to enhance collaboration between DIHs across Europe. The plan is to improve regional DIHs’ services to small and medium-sized (SMEs) manufacturing businesses. These improvements will be made on three different levels:

    1. Policy Impact

    AI REGIO coordinates smart specialisation strategies across European regions and beyond to help scale innovations to global markets. This builds on the Four Motors for Europe movement and the I4MS Community and Innovation Collaboration platform, fostering closer cooperation.

    1. Technological Impact

    The AI REGIO uses previous EU-funded projects like L4MS and AI4EU to create Digital Manufacturing Platforms. These platforms are integrated into the services offered by Digital Innovation Hubs. As a result, Digital Manufacturing Platforms receive support for their business and social impact challenges, while also expanding their AI-enabled technological assets.

    1. Business Impact

    In AI REGIO, both DIHs and SMEs collaborate on 30 regional application experiments that use AI technology. These experiments focus on helping SMEs adopt AI and include considerations for skill development, privacy, and sovereignty preservation.


    The FZI Research Centre for Information Technology is a non-profit organisation that stands for applied research and technology transfer. Since its foundation in 1985, it has been conducting research in various fields related to computer science. Under the motto “Research is the engine of the future”, the institute claims to work towards responsible future shaping.