Siegfried Mercelis tenure track asst. professor - guest prof. external funds

Research Siegfried Mercelis

Research team

Expertise

Hybrid artificial intelligence for modelling and optimization of industrial processes to improve their sustainability. Distributed artificial intelligence in the context of smart mobility and logistics (e.g. automotive, smart shipping, smart traffic).

RELIC 2.0: REinforcement Learning for Interpretable Chemical optimization – phase II. 01/01/2025 - 31/12/2025

Abstract

This project is a continuation of IOF-POC RELIC, aimed at improving the adaptability of Reinforcement Learning (RL) and explainability techniques for chemical process control. Leveraging the results of DAP2CHEM and the improvements from RELIC, we demonstrated a 20-35% increase in process efficiency at production scale proving the technology's viability beyond the pilot phase. In this next phase, RELIC 2.0, the focus will be on enhancing the RL agent to handle multi-objective reward functions, enabling dynamic adaptation to various goals such as time and solvent use. We will also develop a User Interface as a frontend to our existing backend, integrating both to create a Minimum Viable Product (MVP) that will serve as the foundation for a future Software-as-a-Service (SaaS) offering. For valorization, the project will focus on validating the MVP, gathering feedback, and validating market demand through Voice-of Customer analyses. The business model will be continuously refined to align with the evolving needs of customers, with the ultimate objective of laying the groundwork for a successful spinoff launch.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Forecasting and optimization of rail planning using deep learning. 01/05/2024 - 30/04/2028

Abstract

In today's logistics, rail transport progressively shows competitiveness in connecting deep-sea port with the hinterland. As an example, Port of Antwerp-Bruges - the second largest port in Europe – attempts to double the cargo throughput via this modality to 15% by 2030 [1]. Capability to carry large volume, various types of freight, along with high reliability and sustainability are making trains advantageous over traditional trucks. The recent boom of freight railways transport has, however, put more pressure on both inland and deep-sea port infrastructure. In fact, many rail hubs at seaport are still operating sub-optimally due to the incapacity of maximizing their resources utilization. Subsequently, while some bundles are frequently overloaded, others remain rarely used for months. Certain long tracks are blocked to park single wagons during hours, causing shortage when long wagon compositions require service. Train path reservation slot remains fixed for all wagon moving tasks (e.g., 8 hours), resulting in struggles in this resource allocation. These shortcomings are originated from the fact that the rail resource planning is being conducted in a first-come-first-serve, random-pick-up and manual fashion, without having insights of wagon flows in the near future. As a result, even owning large-scale rail resources, some ports still face unworthy shortage or serious delay, which finally adds overheads in total transport cost. Facing this, the optimization of resource allocation based on prediction of up-coming cargo flow will foster the rail infrastructure management, and thus the overall rail operation efficiency. This PhD proposal researches wide range of Machine Learning models to enhance the end-to-end visibility of wagon journey to the deep-sea port. These models forecast the complete path of wagon, from the moments when the long-haul trains are still hours before arrival. The most crucial stages include: arrival time at the main hub, service delay and service time at the shunting yard, train path to move wagon to bundle, public track and time slot to park each wagon, bundling dwelling (such as electric – diesel locomotive shift, which necessitates locomotive and its path allocation), and terminal service (loading/unloading). Next step, the insights learnt from these predictive indicators will be then act as outputs of Optimization phase, to propose the planning of rail tracks, train path, shunting yards and terminal slots which will avoid future shortage, mitigate idle time, and maximize the served cargo volume. Various optimization methods (traditional vs combinatorial neural, single vs multiple objective) will be benchmarked for better accuracy – computational complexity trade-off. The PhD work will be linked with regional (Flanders), national (Belgium) or European research projects, in order to validate the proposed solutions on a real use case with large-scale data. Moreover, it creates synergies upon the solid background of IDLab-imec, which have been showcased through previous projects, including data-driven models, simulation, and optimization aspects, for expanding their contribution in logistics domain.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Safer model-based reinforcement learning for motion planning in autonomous inland shipping. 01/11/2023 - 31/10/2025

Abstract

Currently, there is much research in the field of autonomous navigation. More recently, reinforcement learning (RL) is showing promising results in that field. A type of RL that shows great potential, called model-based RL has some considerable advantages over its model-free counterpart. Notably, it shows potential for safety improvement. Safety is one of the most important challenges that RL in autonomous navigation currently faces.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Goal-Oriented Process Control using Constraint-Guided Model-Based Reinforcement Learning. 01/11/2023 - 31/10/2025

Abstract

Due to its strong economic impact, the field of process control has received much research interest over the years. Whilst traditional control methods have been used in the industry for decades, the application of Machine Learning (ML) has not been properly assessed. An interesting novel field withing ML is Reinforcement Learning (RL), which has repeatedly improved the state-of-the-art (SOTA) in the control of complex systems. Consequently, applying this technique to industrial process control has the potential of strongly improving process efficiency. On the one hand, this leads to reduced cost, resource usage and energy requirements for some of the biggest industries worldwide. On the other hand, this opens a new avenue for collaboration between academics and industry. This project aims to research techniques that are centered around applying RL to industrial process control by developing goal-oriented agents that effectively capture the expectations of the user. (1) An agent with an accurate latent world model will be developed with SOTA performance and strong reasoning capabilities. (2) This agent is extended with a reverse imagination model to reconstruct physical states from latent states. State constraints are applied to these physical states based on expert knowledge to create an intuitive framework for guiding the agent. (3) The agent is then transferred from simulation to reality using offline data to align the internal world model with the real-world environment

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Hybrid AI for Predictive Road Maintenance (HAIRoad). 01/10/2023 - 30/09/2025

Abstract

The current approach to monitoring road quality is based on manual inspections and is labor intensive and relatively expensive. Hybrid AI for Predictive Road Maintenance (HAIRoad) aims to use (hybrid) AI to map the condition of the road network and make recommendations for road maintenance. An efficient and robust data pipeline will be developed using MLOps tools, which allow easy switching between model development and implementation/production. Three demonstrators will illustrate the feasibility of the approach: one with the Port of Antwerp Bruges and two at the municipal level. The demonstrators will allow to validate both the more technical aspects and the market potential. HAIRoad will deliver several innovations such as automated detection of the road conditions, new indicators for road management, sensor fusion by combining information from multiple sensors, and the application of hybrid-AI where we will incorporate physical models of road degradation into data-driven machine learning models.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Data-efficient hybrid modelling for end-point prediction in scaleup of pharmaceutical unit operations. 01/08/2023 - 31/07/2027

Abstract

The project aims to research data-efficient and scalable grey-box modelling for end-point prediction of unit operations in pharmaceutical production processes. We will research a method that makes use of physics-informed neural networks and few-shot learning to achieve this. To enable broader applicability, we will investigate how to efficiently design and calibrate the models in a real-world setting. This method will yield a thorough understanding of the process state during each individual unit operation and provide a twofold benefit for Janssen Pharmaceutica: (1) increase efficiency and decrease cycle times of commercialized processes, and (2) deliver the Best Process At Launch (BPAL) for New Product Introductions (NPIs).

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Strengthening the capacity for excellence of Slovenian and Croatian innovation ecosystems to support the digital and green transitions of maritime regions (INNO2MARE). 01/01/2023 - 31/12/2026

Abstract

The main goal of INNO2MARE is to strengthen the capacity for excellence of Western Slovenian and Adriatic Croatian innovation ecosystems through a set of jointly designed and implemented actions that will support the digital and green transitions of the maritime and connected industries. Based on an in-depth mapping of the ecosystems and needs & gaps analysis, the consortium will formulate a long-term R&I strategy aligned with regional, national and EU strategies, as a visionary framework, and a joint action & investment plan, with concrete steps for building coordinated, resilient, attractive and sustainable maritime innovation ecosystems. To support the joint strategy and provide a model for the future collaborative R&I of the ecosystems' actors, the project will implement three R&I pilot projects that address some of the key challenges related to maritime education and training, security & safety in marine traffic as well as energy conversion and managementsystems' efficiency. These pilots will be the basisfor further development,scale-up and translation of the generated research results into innovative business opportunities through the coordinated mobilisation of public and private funding. The consortium will also implement innovative programmes that will support the engagement of citizens in the innovation processes, knowledge transfer for mutual learning, entrepreneurship & smart skills training and attraction of best talents, involving more than 1.000 participants across the Quadruple Helix. In all the project activities, the two ecosystems will strongly benefit from the sharing of best practices of the Flemish ecosystem, one of the most developed maritime innovation ecosystems globally. The project will contribute to reducing the innovation divide in Europe by systematically connecting the innovation actors within and between the ecosystems and creating synergies in R&I investments' planning and execution, thus developing a true innovation culture

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Accident-prone Vision-based Simulation for Autonomous Safety-critical Systems 01/11/2022 - 31/10/2025

Abstract

Autonomous navigation has been gaining much traction recently. As a result, we see autonomy developing in vehicles and finding its way in many transportation sectors (including smart shipping). Nevertheless, the current state-of-the-art (SOTA) technology is not mature enough to have a widespread application at a higher autonomy level (e.g. level 4 and above). The main reason is that these systems are trained on a lot of real-world data, which often lacks accident-prone scenarios. In order to solve this problem, I propose a solution based on data-driven neural simulations that provide realistic data based on real-world samples and generate unsafe scenarios (collisions, accidents, etc.). Moreover, my system also provides safety checks to validate unsafe scenarios and provide safe boundaries for the current autonomous systems.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

IMEC-Driving the future of water resource recovery facilities through data intelligence (DARROW). 01/09/2022 - 31/08/2026

Abstract

The wastewater sector is going through a profound transformation with energy efficiency and resource recovery as key priorities in wastewater treatment plants (WWTP) and these installations started to be perceived as Water Resource Recovery Facilities (WRRF). Under this context, the exploitation of data through artificial intelligence tools with the objective of accelerating the transition of WWTP to WRRF has not been fully addressed yet. When compared to treatment technologies, the deployment of AI-powered tools in production is much faster and, therefore, provides immediate benefits. In that sense, three main barriers have been identified in this domain: i) Mechanistic mathematical models involve complex formulations and specific terminology that are difficult to understand for plant operators; ii) WRRF are harsh environments with strong impact on the quality of data; iii) Essential information in WRRFs is limited and not continuously available. In particular, to overcome these challenges, DARROW will build and demonstrate into an operational environment, an innovative, optimised, modular, and flexible data-driven AI solution to make existing WWTP more autonomous, more energy efficient and better prepared for their transformation into WRRF. DARROW will take advantage of existing AI & Data analysis techniques with the final objective of contributing to a greener planet by: i) Reducing energy consumption of WRRF; ii) Reducing Greenhouse Gas Emissions of WRRF; iii) Increasing Resource Recovery iv) Improving water quality.To do so, DARROW gathers the necessary experience, knowledge and resources through a multi-stakeholder approach that covers the whole value chain of the project. It consists of a multidisciplinary team of 8 entities from 4 different EU countries (Spain, Belgium, Germany and Netherlands), among which, 3 RTOs, 1 university,1 NPO, and 3 SMEs to ensure market exploitation (2 industrial companies and 1 water company).

Researcher(s)

Research team(s)

Project website

Project type(s)

  • Research Project

IMEC-AI Pathfinder. 01/01/2022 - 31/12/2025

Abstract

The overall goal of the AI PathFinder project is to support Flemish food companies to develop their AI strategy and thus accelerate concrete AI adoption. This will eventually make them a company that is much more strongly armed for the future, thus maintaining competitiveness and growth. The focus is therefore on inspiring, encouraging and facilitating adoption of AI-based solutions for concrete challenges, needs and opportunities of food companies.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Knowledge Based Neural Network Compression: Context-Aware Model Abstractions 01/11/2021 - 31/10/2025

Abstract

In the state-of-the-practice IoT platforms complex decisions based on sensor information are made in a centralized data center. Each sensor sends its information over thereafter a decision is send to actuators. In certain applications the latency imposed by this communication can lead to problems. For this, decisions should be made on the edge devices themselves. This is what the research track on resource and context aware AI is about. We want to develop edge inference systems that dynamically reconfigure to adapt to changing environments and resources constraints. This work is focused on compressing neural networks. In this work we want to extend on the current state-of-the-art on neural network compression by incorporating a knowledge-based pruning method. With knowledge based we mean that we first determine the locations of specific task related knowledge in the network and use this to guide the pruning. This way we can make the networks adjustable to environmental characteristics and hardware constraints. For some tasks in a specific environment, it might be favorable to reduce the accuracy of certain classes in favor of resource gain. For example, the classification of certain types of traffic sign types can be less accurate on highways than in a city center. Based on these requirements we want to selectively prune by locating specific task related concepts. By removing them we expect to achieve higher compression ratios compared to the state-of-the-art.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Goldilocks' Fusion: Adaptive and Robust Sensor Fusion in Resource-Constrained Robotic Systems. 01/11/2021 - 31/10/2025

Abstract

In recent years, autonomous robotic systems have gained lots of attention from the academic world and industry. The many applications in industrial fields going from manufacturing, mining and surveillance makes the study on autonomous systems interesting with lots of valorization potential. The cost of these autonomous systems is currently extremely high as expensive computational platforms and sensors suites are used to provide necessary levels of safety and autonomy. Using the measurements from different sensors, an environment representation is created to make navigational decisions. While the environment representation determines the complexity of the behavior that can be achieved, the detail stored in this representation is dependent on the available computational resources and sensor data. The goal of this research project is to enable an autonomous agent to select the optimal heterogenous set of sensors to create an environment representation of the appropriate complexity for the current situation. Resource awareness plays an important role in our research as we aim to reduce computational workloads on the autonomous vehicles, which means less expensive computational platforms can be used. Additionally, increased reliably and accuracy in environment perception will benefit the autonomy of these systems. Less expensive autonomous systems while being efficient in the use of resources will benefit and increase the adoption of autonomous vehicles.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Distributed multi-modal data fusion using graph-based deep learning for situational awareness in intelligent transport systems. 01/11/2021 - 31/10/2025

Abstract

Reliability and accuracy are the two fundamental requirements for intelligent transport systems (ITS). The reliability of active perception for situational awareness algorithms has significantly improved in the past few years due to AI developments. Situational awareness can be improved through exchange of information between multiple agents. Making it complex to accomplish high accuracy at low computational cost cooperatively is critical to ensuring safe and reliable transport systems. This research will tackle the main challenges for shared situational awareness that requires perception from multiple sensor streams and multiple agents. This research will tackle the local sensor fusion problem with graph-based deep learning. Local sensor fusion is the fusion at the agent level where multiple mounted sensors will be used to solve a defined task. By exploiting the structural information in multiple modalities, the proposed solution will construct graph-based deep learning. Then distributed fusion will be accomplished by fusing predictions from multiple agents. As a result, the predictions can be fused across multiple agents to produce a richer situational awareness. The advantage of doing distributed fusion is evident in situations where a single agent's perception is not enough. This will be achieved by modeling spatio-temporal graph networks and studying dynamic updates in the graphs. The results will be validated using real-life benchmark datasets and simulation engine.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Knowledge Based Neural Network Compression: Quality-Aware Model Abstractions. 01/10/2021 - 30/09/2025

Abstract

In the state-of-the-practice IoT platforms complex decisions based on sensor information are made in a centralized data center. Each sensor sends its information over thereafter a decision is send to actuators. In certain applications the latency imposed by this communication can lead to problems. In real time applications it is crucial for the decision to be taken immediately. For this complex decisions should be made on the edge devices themselves. This is what the research track on resource and context aware AI is about. In this we want to develop inference edge systems that dynamically reconfigure to adapt to changing environments and resources constraints. This work if focused on compressing AI processing blocks, specifically neural networks. In this work we want to extend on the current state-of-the-art methods on neural network compression by incorporating a knowledge-based pruning method. By knowledge based we mean we want to prune a neural network in a context aware manner. A certain application context will impose requirements of the outputs of the network. For example, on a highway is the detection of pedestrians less important than cars. Based on these requirements we want to selectively prune a network by locating knowledge concepts related to the outputs. By selectively pruning them we expect to achieve higher compression ratios compared to the state-of-the-art for context specific networks.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

IMEC- Portable innovation open network for efficiency and emissions reduction solutions (PIONEERS). 01/10/2021 - 30/09/2025

Abstract

PIONEERS brings together four ports with different characteristics, but shared commitments towards meeting the Green Deal goals and Blue Growth socio-economic aims, in order to address the challenge for European ports of reducing GHG emissions while remaining competitive. In order to achieve these ambitions, the Ports of Antwerp, Barcelona, Venlo and Constanta will implement green port innovation demonstrations across four main pillars: clean energy production and supply, sustainable port design, modal shift and flows optimization, and digital transformation. Actions include: renewable energy generation and deployment of electric, hydrogen and methanol vehicles; building and heating networks retrofit for energy efficiency and implementation of circular economy approaches in infrastructure works; together with deployment of digital platforms (utilising AI and 5G technologies) to promote modal shift of passengers and freight, ensure optimised vehicle, vessel and container movements and allocations, and facilitate vehicle automation. These demonstrations form integrated packages aligned with other linked activities of the ports and their neighbouring city communities. Forming an Open Innovation Network for exchange, the ports, technology and support partners will progress through project phases of innovation demonstration, scale-up and co-transferability. Rigorous innovation and transfer processes will address technology evaluation and business case development for exploitation, as well as creating the institutional, regulatory and financial frameworks for green ports to flourish from technical innovation pilots to widespread solutions. These processes will inform and be undertaken in parallel with masterplan development and refinement, providing a Master Plan and roadmap for energy transition at the PIONEERS ports, and handbook to guide green port planning and implementation for different typologies of ports across Europe.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

IDLab - Internet and Data Lab 01/01/2021 - 31/12/2026

Abstract

The IOF consortium IDLab is composed of academic supervisors at the IDLab Research Group, a UAntwerp research group with members from the Faculty of Science and the Faculty of Applied Engineering. IDLab develops innovative digital solutions in the area of two main research lines: (1) Internet technologies, focusing on wireless networking and Internet of Things (IoT), and (2) Data science, focussing on distributed intelligence and Artificial Intelligence (AI). The mission of the IDLab consortium is to be the number one research and innovation partner in Flanders and leading partner worldwide, in the above research areas, especially applied in a city and its metropolitan surroundings (industry, ports & roads). To realize its mission, IDLab looks at integrated solutions from an application and technology perspective. From an application point of view, we explicitly provide solutions for all stakeholders in metropolitan areas aiming to cross-fertilize these applications. From a technological point of view, our research includes hardware prototyping, connectivity and AI, enabling us to provide a complete integrated solution to our industrial partners from sensor to software. Over the past years, IDLab has been connecting the city and its surroundings with sensors and actuators. It is time to (1) reliably and efficiently connect the data in an integrated way to (2) turn them into knowledgeable insights and intelligent actions. This perfectly matches with our two main research lines that we want to extensively valorise the upcoming years. The IDLab consortium has a unique position in the Flemish eco-system to realize this mission as it is strategically placed across different research and innovation stakeholders: (1) IDLab is a research group embedded in the Strategic Research Centre imec, a leading research institute in the domain of nano-electronics, and more recently through groups such as IDLab, in the domain of digital technology. (2) IDLab has a strategic link with IDLab Ghent, a research group at Ghent University. While each group has its own research activities, we define a common strategy and for the Flemish ecosystem, we are perceived as the leading partner in the research we are performing. (3) IDLab is the co-founder of The Beacon, an Antwerp-based eco-system on innovation where start-ups, scale ups, etc. that work on IoT and AI solutions for the city, logistics, mobility and industry 4.0 come together. (4) Within the valorisation at UAntwerp, IDLab contributes to the valorisation within the domain 'Metropolitanism, Smart City and Mobility'. To realize our valorisation targets, IDLab will define four valorisation programs: VP1: Emerging technologies for next-generation IoT; VP2: Human-like artificial Intelligence; VP3: Learning at the edge; VP4: Deterministic communication networks. Each of these valorisation programs is led by one of the (co-)promoters of the IDLab consortium, and every program is composed of two or three innovation lines. This way, the IDLab research will be translated into a clear program offer towards our (industrial) partners, allowing us to build a tailored offer. Each valorisation program will contribute to the different IOF objectives, but in a differentiated manner. Based on our current experience, some valorisation programs are focusing more on local partners, while others are mainly targeting international and EU funded research projects.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Learning to communicate efficiently with multi-agent reinforcement learning for distributed control applications. 01/11/2020 - 25/02/2025

Abstract

In recent years, there has been increased interest in the field of multi-agent reinforcement learning. For tasks where cooperation between agents is required, researchers are looking towards techniques to allow the agents to learn to communicate while simultaneously learning how to act in the environment. Current state-of-the-art techniques often use broadcast communication. However, this is not scalable to real world applications. Therefore, I want to develop methods to make this communication more efficient. The goal of this research project is to reduce the amount of messages that are sent, while still maintaining the same performance. To reach this goal, I will look at techniques to communicate with a variable amount of agents, at techniques to limit communication using relevance metrics and signatures and at techniques to encourage hopping behavior in agents. The methods proposed in this research project are essential to be able to create scalable control applications by distributing them in combination with scalable learned communication. The developed methods will be validated on simulations of traffic light control.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

RELIC: "REinforcement Learning for Interpretable Chemical optimization". 01/09/2023 - 31/08/2024

Abstract

The pharmaceutical and chemical industries are perpetually challenged by the need to enhance efficiency, scalability, adaptability, and sustainability in their processes, all while adhering to stringent regulations. It is here that artificial intelligence (AI) emerges as a promising solution, offering innovative and efficient means to navigate these complexities. At IDLab, our expertise in employing AI as a problem-solving tool in these sectors has been marked by significant achievements, particularly through our Catalisti-ICON project, DAP2CHEM. We successfully demonstrated the potential of Reinforcement Learning (RL) and Explainable AI (XAI) to improve operational efficiency by 25-35% and resulted in reduced energy and raw material consumption, further enhancing sustainability. Our technology has made considerable strides in two critical areas of AI application in high-risk industries: maintaining operational safety and ensuring transparency, while significantly increasing the efficiency of the process. We have utilized RL in a safety-conscious manner, ensuring that operations are reliable and remain within established safety parameters. At the same time, our work with XAI has yielded humanunderstandable explanations for AI decisions, significantly enhancing the transparency of our technology. Building upon these achievements, the goal of this project is to increase the adaptability and scalability of our technology across different operations and production scales. We also aim to enhance training efficiency, increase automation, and make XAI explanations more intuitive for users with various levels of expertise. This direction aligns perfectly with our ongoing commitment to satisfy a wider range of industry needs. The improved transferability and scalability of the technology will increase the level of commercial readiness in order to valorize it in the Flemish and international chemical and pharmaceutical sector. A substantial component of our roadmap is the valorization of the technology. The project aims to enhance its commercial readiness by ensuring the technology is adaptable, scalable, and user-friendly, catering to the wide-ranging needs of the chemical and pharmaceutical industries. Recognizing a promising market demand, the formation of a spin-off company is a viable consideration, with the potential to provide specialized AI services in these sectors. Aiming to be a key player in the digital transformation of the chemical and pharmaceutical industries, we envision a future where our advanced AI solutions become integral to their operational efficiency and sustainability.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

QPM Application Workloads (Budget IMEC.Invest). 01/01/2023 - 31/12/2023

Abstract

The objective is the creation of advanced hybrid modelling techniques for computational fluid-dynamics and evaluating their accuracy and computational workload compared to advanced first-principles models.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Goal-Oriented Process Control by Including Expert Knowledge in Model-Based Reinforcement Learning using Soft Constraints. 01/11/2022 - 31/10/2023

Abstract

Due to its strong economic impact, the field of process control has received much research interest over the years. Whilst traditional control methods have been used in the industry for decades, the application of Machine Learning (ML) has not been properly assessed. An interesting novel field withing ML is Reinforcement Learning (RL), which has repeatedly improved the state-of-the-art (SOTA) in the control of complex systems. Consequently, applying this technique to industrial process control has the potential of strongly improving process efficiency. On the one hand, this leads to reduced cost, resource usage and energy requirements for some of the biggest industries worldwide. On the other hand, this opens a new avenue for collaboration between academics and industry. This project aims to research techniques that are centered around applying RL to industrial process control by developing goal-oriented agents that effectively capture the expectations of the user. (1) An agent with an accurate latent world model will be developed with SOTA performance and strong reasoning capabilities. (2) This agent is extended with a reverse imagination model to reconstruct physical states from latent states. State constraints are applied to these physical states based on expert knowledge to create an intuitive framework for guiding the agent. (3) The agent is then transferred from simulation to reality using offline data to align the internal world model with the real-world environment.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

OptiRoutS: private routing service that proactively contributes to meeting public mobility goals. 01/10/2022 - 30/09/2024

Abstract

Authorities have grown concerned over the negative impact of in-car routing services on smooth, safe and green mobility, as these often fail to consider the social cost inherent to the usage of the road by their users. Private partners (such as intermediaries, end-user service providers or mobility consultants) are increasingly involved in rectifying the worrisome aspects of these routing services, though still face the challenge on how to go from public mobility goals to impactful policy advice or route guidance. The key aspects of this challenge are the lack of: (i) a large-scale and robust methodology to quantify the social cost of traffic on a road network; and (ii) academic knowledge on how to implement impactful routing advice, e.g., via altruistic rewards. In OptiRoutS, three industrial partners (Be-Mobile, Movias and TML) and one public partner (AWV) team up with four academic partners (IDLab-Antwerp, IDLab-Ghent, CIB-KUL and SMIT-VUB) to address these challenges and build services that contribute to smoother, safer and more sustainable mobility.  The innovations in OptiRoutS will strengthen the partners positions in two promising markets – traffic policy support and interactive traffic management, thus providing significant scope for valorization. 

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Chemistry & Hybrid AI - Boosting efficiency of product development, monitoring, analysis and production processes in chemistry by leveraging expert knowledge through Hybrid AI (CHAI). 01/03/2022 - 29/02/2024

Abstract

CHAI will transform chemical process control and stability analysis by fully embracing the vision of hybrid AI. In CHAI, smart AI tools are developed for next-generation chemical process control, by explicitly coding expert knowledge into hybrid AI models. Currently, process engineers need many years of experience and training before they can tackle the most challenging problems completely independently. By embracing hybrid AI, process engineers and operators can be empowered by giving each of them access to (1) everyone's experience and expertise, and (2) a smart system that performs outcome predictions and proposes control actions. The CHAI consortium expects efficiency gains and a better understanding of product characterization; CHAI aspires to influence R&D efficiency and production through cost reduction, and to eliminate the risk of recalling millions of euros of products that do not meet quality standards.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

IMEC-Enhanced data processing techniques for dynamic management of multimodal traffic (TANGENT). 01/09/2021 - 31/08/2024

Abstract

The European transport faces major challenges in terms of safety, greenhouse gas emissions, traffic congestion and its derived costs. In addition, the development of disruptive technologies and emergence of new mobility solutions generate a revolution in transport network and traffic management. In this context, TANGENT aims to develop new complementary tools for optimising traffic operations in a coordinated and dynamic way from a multimodal perspective and considering automated/non-automated vehicles, passengers and freight transport. TANGENT will research on advanced techniques on modelling and simulation, such as prediction and simulation models for future demand & supply of transport; optimisation techniques for balancing the demand flows between the means of transport; and users travel behaviour modelling. As result, a set of applications for decision-making support will be delivered creating a framework for coordinated traffic and transport management, encompassing an enhanced mobility information service and dashboard with associated APIs and advanced functionalities with a two-fold approach: to provide real-time traffic management recommendations and to support Transport Authorities to design network-wide optimal strategies. The framework also aims at supporting a multi-actor cooperation approach for transport network management by enabling communication channels. In this way, the services target to different actors in traffic management. The results will be tested in three case studies: Rennes (FR), Lisbon (PT), Great Manchester (UK) and a virtual case study in Athens (HE)with real data from various modes of transport, under different traffic events such as bottlenecks, accidents, pedestrian flow etc. The impact will be assessed to reach expected reduction targets of 10% in travel time, 8-10% in CO2 emissions, 5% of accidents, 5-10% increase in use public transport and use of active modes or 10% of economic costs due to a more efficient management.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

AI For Food Logistics aims to achieve a highly reliable, just-in-time delivery experience for fresh food through end-to-end optimization of the logistics chain (AI4FoodLogistics). 01/09/2021 - 31/08/2023

Abstract

Demand forecasting in retail still suffers from the so called bullwhip effect: a small change in point-of-sale demand can cause a large fluctuation in demand at the wholesale, distribution center and supplier. For the food retail chain this materializes into (i) the inability to cope with unexpected events or (ii) to further reduce food waste, and (iii) a weaker position towards e-commerce. AI4FoodLogistics aims to tackle these challenges - focusing on fresh food delivery - by addressing key shortcomings of current tactics. A highly reliable, just-in-time delivery experience for fresh food is targeted, that leverages a novel data architecture capable to propagate data across the value chain in a more scalable, cost-effective way. The consortium spans the full value chain, from farm to fork, and will focus on advancing state-of-the art technology to increase the trustworthiness of demand forecasting, logistics scheduling and personalized recommendations. Key objectives of AI4FoodLogistics are lowering overall logistics cost (ca. 13M euro/year) decreasing food waste in Flanders (at least 15M euro/year), and increasing the share of locally produced healthy food. The outcome will be validated by combining a simulator with in-the-field-validated data-driven models.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

IMEC-Update and maintain the AI model. 01/07/2021 - 31/12/2023

Abstract

Update and maintain the AI model every quarter. imec project on recycling electronic waste automatically instead of manually by using an AI model. The AI model was developed in a previous project and this project is for the maintenance of the model and making the necessary updates.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Augmenting clinical decision-making processes for primary care physiotherapists based on state-of-the-art artificial intelligence and deep learning techniques. 01/07/2021 - 31/12/2022

Abstract

The objective of this project is to demonstrate how patient-related physiotherapy data that is collected and stored in a structured way, can be used for data-analysis by the use of deep learning, a part of a broader family of machine learning methods. Neural networks will be used to determine which therapeutic approach can be best used for what type of patient to increase physical activity, thereby demonstrating the feasibility of using data-analysis to develop effective therapeutic strategies in patients with cardiorespiratory and metabolic diseases. Demonstrating the feasibility of data gathering, storage and analysis in physiotherapy in internal diseases in a primary care setting will be a first and major step in developing data-driven therapy. The results of this project will facilitate and enable further research in the development of data-driven medicine in multiple diseases, as well as the development of data-monitoring and tele-coaching application in healthcare. The combined expertise of both research groups, in partnership with the Belgian Physiotherapy Association (Axxon), allows this consortium to take a head start in data-driven physiotherapy research and to become a pioneer in this field in Europe.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Visualising 'material spatial dimension of waste flows' in the province of Antwerp (ATLANTES). 01/06/2021 - 31/05/2023

Abstract

Flanders and the Province of Antwerp aim to reduce the footprint of raw material consumption by 30% towards 2030, focusing more on circular economy (CE). This project aims to support the CE approach by developing an online platform that visualizes the waste flows from and to the province of Antwerp. After all, thanks to the analysis of waste flows, governments and public institutions can maximize the impact of their policy choices and the results achieved in this CE transition. In addition, it enables companies to trace their waste materials at a provincial level, but also to develop new production processes in which they also reuse their waste materials.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Using Model-Based Reinforcement Learning combined with Monte-Carlo Tree Search to optimize Neural Networks for Embedded Devices. 01/11/2020 - 31/10/2024

Abstract

Currently, most AI systems are being run in cloud environments. For some systems, like real-time systems, this can be troublesome, and moving these AI algorithms to the edge can provide a solution to these problems. The aim of my research is to use reinforcement learning techniques to design neural networks with performance rivalling that of modern, state-of-the-art systems, while reducing the resource consumption of these systems to a level that is manageable for edge devices. In order to achieve this goal, my work is split into 3 large components: multi-objective optimization, hardware embeddings and model-based reinforcement learning (MBRL) using monte carlo tree search (MCTS). The first component of my research, will deal with the scalarization of a multi-objective reward function, into a scalar reward. This is necessary for reinforcement learning systems, since they take a single reward value as feedback. For the second component of my research I will try to find a way to represent a certain piece of hardware, in a neural-network friendly manner. This is necessary for our system to be able to be able to exploit the architectural features of a specific piece hardware. Finally, I will introduce MBRL using MCTS to the field of neural architecture search. In this component, I will utilize the developed scalarization techniques and hardware representation from the first two components and a MBRL system to generate neural network architectures targeted at specific devices.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Sustainable Internet of Batteryless Things (IoBaleT). 01/10/2020 - 30/09/2024

Abstract

The Internet of Things (IoT) vision has enabled the wireless connection of billions of battery-powered devices to the Internet. However, batteries are expensive, bulky, cause pollution and degrade after a few years. Replacing and disposing of billions of dead batteries every year is costly and unsustainable. We posit the vision of a sustainable Internet of Battery-Less Things (IoBaLeT). We imagine battery-less devices storing small amounts of energy in capacitors, harvested from their environment or obtained through simultaneous wireless information and power transfer (SWIPT). Using this energy, these intermittently-powered devices are able to cooperatively perform sensing, actuation and communication tasks. Existing battery-less technology has many shortcomings. Such devices, usually based on passive RFID and backscatter, only support simple sensing, unable to handle more complex application logic. Networks do not scale, have a short range and a very low throughput. The goal of IoBaLeT is to bring battery-less technology to the next level. We envision battery-less devices and networks that support complex sensing and actuation applications, and offer throughput, scalability and range on-par with their battery-powered counterparts. To achieve this, we propose a novel battery-less IoT device design that relies on a combination of SWIPT, hybrid energy harvesting, active transmissions and wake-up radios. The project will innovate in terms of SWIPT efficiency, battery-less networking protocols, and distributed intermittent computing paradigms and scheduling algorithms. Leaving batteries behind will enable IoT applications at an unprecedented scale, with a significantly extended lifetime and in hard-to-reach places.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Time-Sensitive Computing on Battery-Less IoT Devices 01/01/2020 - 31/12/2021

Abstract

The Internet of Things (IoT) is largely powered by batteries. This poses significant challenges for its sustainability and longevity, as batteries are short-lived, bulky and polluting. To overcome this problem, we posit the vision of a battery-less IoT network, where devices are powered by energy harvesting and tiny long-lived capacitors. However, such devices often run out of power, resulting in intermittent on-off behavior. Traditional static sequential applications cannot handle such behavior, as they lose forward progress. This problem can be solved with task-based applications, consisting of a chain of interconnected tasks. Each task performs some atomic function, and its output is saved in non-volatile memory after it successfully completes. This allows the application to ensure forward progress in face of frequent power failures. Optimally scheduling the execution of such tasks, in face of the specific behavior of various energy harvesters, as well as the capacitor, and given extremely constrained resources of battery-less devices, is non-trivial. In this project, we propose a novel task scheduler that takes these aspects, as well as the deadline of tasks into account.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Multi-Agent Communication and Behaviour Training using Reinforcement Learning. 01/11/2019 - 31/10/2023

Abstract

Many real-world applications require intelligent cooperative agents that can work together to solve a problem. An example of such an cooperative multi-agent application is the control of multiple autonomous vehicles. Multi-agent reinforcement learning is a wellresearched topic and many solutions exist in the state-of-the-art. Recently, the research community was able to create agents that learn how to communicate with each other to reach their goal. This is a new subfield of the multi-agent reinforcement learning domain in which we will research how we can achieve decentralised training of these communicating agents. This will allow us to create heterogeneous agents that can communicate or continue to train the communicating agents after their deployment which is not possible with the current state-of-the-art methods. In this project, I will extend the state-of-the-art by investigation how we can communicate with an unknown number of other agents which is a problem with state-ofthe- art methods. Next, I will work on the feedback structure that is used to train the communication between the agents. After the feedback structure I will work on splitting the agents architecture to create environment specific agents. I hypothesise that this will decrease the training time of the communication policy which is required for decentralised training. These advancements will be combined to create agents that we can train in a decentralised setting.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Simulation based testing of large scale internet of things applications. 01/01/2019 - 31/12/2021

Abstract

The goal of this project is to introduce a simulation based methodology which will be used to cope with the scalability constraints of modern IoT software testing, and more specifically the testing of ultra large scale systems with emergent behavior. With IoT becoming more mainstream and with the rise in the amount of devices getting interconnected, the complexity and scale of the IoT landscape will largely increase. This interoperability between IoT devices and actuators of all sorts will prove to be vital for future IoT applications. As a result of the increased scale and diversity and because of modern decentralized IoT architectures such as Edge computing, we see that a whole new type of IoT application will gain importance. A type of application where local decentralized interaction between devices and actors will lead to a global emergent behavior. The concept of emergence can be compared to a flock of birds, where local interactions between individual birds lead to a global optimized behavior. This idea is also very relevant in IoT, imagine for example a smart traffic light application where local interactions between traffic lights could lead to a global optimized traffic flow. This type of IoT application will however lead to major difficulties with regards to application validation, testing and calibration. That is because in order for realistic emergent behavior to arise, the IoT application will need to be executed in a large-scale and diverse environment. An environment that resembles the eventual operational environment. Deploying such applications to a real-life isolated IoT testbed would be impractical as the cost of setting up such an environment at a realistic scale is too high and requires too much effort in the early stages of development. Instead of relying on expensive test beds, we propose a large scale simulation based approach. Such a simulation -based system needs to incorporate hundreds of thousands of virtual sensors interacting among each other and with the environment. The behavior of these systems will need to be modeled carefully. However, this leads to additional technical challenges. Also all virtual sensors in the system should be continuously active to interact in a real-time fashion with other systems. That is because an important part of the behavior of conventional IoT systems and EBI systems is controlled by an IoT middle-ware, the simulated entities should be able to interact with the middleware as if they were real-life IoT entities. We refer to this as software-in-the-loop (SIL) simulation. Because of this real-time requirement, a great amount of simulation entities should run in parallel which highly increases the computational complexity. Solely relying on state-of-the-art large-scale simulation techniques is insufficient. The contribution of this project is focused on the creation of a methodology for running real-time, large-scale simulations for testing and analyzing both conventional IoT systems and emergent behavior based IoT systems. We will focus on two major tracks, in the first we will reduce the computational complexity by dynamically increasing abstraction levels of simulation models and in the second track we aim at reducing network communication overhead of distributed simulations by optimizing the partitioning of simulation entities over multiple simulation servers.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

A Methodology for Analysis and Optimization of Distributed Artificial Intelligence. 01/07/2018 - 31/12/2019

Abstract

Although the foundations of Artificial Intelligence (AI) have been around for a long time, advances in computational performance and research in novel AI techniques have led to a revival of this research domain. With the advent of the Internet of Things (IoT), numerous "smart" applications driven by AI, have found their way into our everyday lives. Due to the computational complexity of these techniques, currently a common approach is to minimize the computations performed on the user's device and to perform the bulk of the work in a cloud environment. However, with a foresight of over 20 billion smart devices by 2020, handling this data with a cloud-centric approach cannot be maintained. In order to continue the AI revolution, alternative approaches are needed in which the AI is distributed across devices closer to the edge of the IoT network. Current AI solutions mostly focus on large-scale cloud environments or high performance devices. IoT devices, however, are very diverse in hardware architecture and often constrained in resources. Depending on the hardware and software constraints (e.g. timing requirements, computational, memory and energy constraints), tailored optimization strategies are needed. In order to allow distribution of AI algorithms in such a diverse environment, two gaps in the state of the art need to be bridged. In this research project, we will investigate (1) a systematic analysis method for Artificial Intelligence to determine the characteristics of these algorithms and (2) define a method for optimal distribution of AI in the context of IoT.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project

Optimization and parallellisation of real-time media processing on embedded systems by abstraction of software-and hardwarebehaviour. 01/09/2012 - 31/08/2016

Abstract

This project represents a research agreement between the UA and on the onther hand IWT. UA provides IWT research results mentioned in the title of the project under the conditions as stipulated in this contract.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project