The Innovation Radar showcases breakthrough technologies and solutions developed through European research programmes. Being included means that the innovations have been recognised by the European Commission as having strong technological potential and relevance for industry and society.

The TUPLES project focuses on developing trustworthy, explainable and robust AI solutions for complex planning and scheduling problems across sectors such as logistics, energy, manufacturing, and transportation. The inclusion of 11 innovations in the Innovation Radar highlights the breadth and impact of the work carried out by the consortium.

A diverse portfolio of AI innovations

The recognised innovations span a wide range of domains, including explainable AI for logistics and transport, advanced optimisation algorithms, decision support systems, and industrial AI platforms.

Among the highlighted developments are:

• AI Explainable Planning Routing Support System for transport route diversion during incidents and disruptions.
• Algorithms for compressing tree ensemble to improve the efficiency of machine learning models.
• Explainable AI planning systems for industrial supply chains, enabling intelligent storage planning and logistics decision-making.
• Resource-constrained scheduling AI tools that support workforce allocation and dynamic rescheduling of transport operations.
• IPEXCO, a platform for iterative planning with conflict explanations, improving transparency in automated decision systems.
• Advanced clustering techniques for waste collection routing, improving efficiency and sustainability in urban logistics.
• Robust optimisation methods for heat and power generation plants, supporting more resilient and efficient energy management.
• The Self-Assessment Tool for Trustworthiness in Planning and Scheduling, designed to help organisations evaluate AI systems against key trustworthy AI principles.

From research to real-world impact

Several innovations also demonstrate strong market readiness and real-world deployment potential.

The Competition Platform for AI/ML and decision-making models, has been assessed as Tech Ready with Very High market creation potential. The platform enables industrial and academic communities to tackle complex optimisation and AI challenges through competitions and collaborative experimentation.

Another innovation with Business Ready maturity is the Football Squad Management Decision Support System, which supports data-driven strategic decisions by professional football clubs.

Strengthening Europe’s innovation ecosystem

The strong presence of TUPLES innovations on the Innovation Radar reflects the collaborative effort of leading universities, research centres, and industrial partners across Europe. These results demonstrate how European research collaborations can translate cutting-edge scientific advances into technologies with tangible societal and economic value.

You can explore the innovations directly on the Innovation Radar platform:
https://innovation-radar.ec.europa.eu/

Here is the full list: TUPLES INNOVATION RADAR

In the final days of the TUPLES project, our consortium gathered one last time in person for the Final General Assembly, hosted in Toulouse at the Airbus Saint-Martin site — the very place where the iconic Beluga™ aircraft perform their impressive daily loading and unloading operations.

These same operations were at the heart of the Beluga™ AI Challenge, a competition that invited the international planning and scheduling community to address a real-world industrial problem through trustworthy AI.

The Award Ceremony, introduced by Romaric Redon, COO of ANITI Toulouse, and Florent Teichteil-Koenigsbuch, AI Decision-Making Expert @1XRD Airbus, celebrated the outstanding solutions developed by the two winning teams:

• Bernardino Romera Paredes (Hiverge), winner of the Scalability Track
• Daniel Gnad and Elliot Gestrin (Linköping University), winners of the Explainability Track

Before the event, the winners and consortium members had the chance to witness the Beluga™ in action. Seeing the aircraft’s impressive scale and the complexity of its loading operations made it even more compelling to later hear directly from the winners about their innovative approaches to this challenging planning problem.

It was a truly special moment for the TUPLES partners — an inspiring closure to three years of collaboration, research, and innovation towards Trustworthy Planning and Scheduling.

In the coming weeks, we will publish all the final outcomes of the project on our website.
For now, we invite you to watch the recording of the Award Ceremony below.

Watch the full ceremony here

TUPLES has focused on building trustworthy planning and scheduling systems that are scalable, safe, robust, and explainable—bridging fundamental research with impactful applications.

To achieve these objectives, the cornerstones of our scientific contributions were:

(1) combining symbolic P&S methods with data-driven methods to benefit from the scalability and modeling power of the latter, while gaining the transparency, robustness, and safety of the former.

(2) developing rigorous explanations and verification approaches for ensuring the transparency, robustness, and safety of a sequence of interacting machine learned decisions.

From a practical standpoint, the project demonstrated and evaluated our novel and rigorous methods in a laboratory environment, on a range of use-cases in manufacturing, aircraft operations, sport management, waste collection, and energy management.

During this period, from a scientific standpoint, we continued to:

• advance the state of the art with novel hybrid approaches (model-based / data-driven) for planning and scheduling that aim at increasing the robustness and/or scalability of existing methods;
• developed new methods for verifying, testing, improving or enforcing the safety and robustness of the muti-step decisions recommended by these approaches;
• came up with interactive explanation approaches allowing users of planning and scheduling systems to understand why a solution was recommended over others, and to use this understanding to guide the system towards a solution satisfying their preferences.
  
  

Altogether, the project produced 68 scientific publications, over 50 of which were accepted at top-tier AI conferences and journals, with two receiving awards. Many of these contributions were accompanied by publicly released code, ensuring accessibility and reusability by the broader community.

These results were made possible thanks to the support of the European Union, whose Horizon Europe funding has been instrumental in advancing both the scientific and practical outcomes of the project, while uniting teams from different countries and disciplines in pursuit of common goals.

All publications are available in the dedicated section of the TUPLES website, where each entry directly links to the corresponding paper.

This resource provides open access to the knowledge generated during the project and serves as a lasting legacy of the scientific advances achieved.

A Challenge at the Frontiers of Industrial AI

The Beluga™ AI Challenge was designed to test cutting-edge approaches to some of the most demanding problems in industrial planning and scheduling. The challenge featured three tracks, reflecting two complementary facets of Trustworthy AI: Scalability and Explainability.

The Scalability facet, in particular, was divided into two distinct challenges. The deterministic track focused on traditional approaches where planning models operate under fixed conditions, while the probabilistic track encouraged the exploration of techniques able to cope with uncertainty, such as reinforcement learning and related methods.

Hiverge’s Outstanding Achievement

We are pleased to highlight the outstanding achievement of Bernardino Romera-Paredes and Kerry He from Hiverge, who won both Scalability tracks. Their solutions addressed the complexity of optimizing assembly and transport schedules in the Beluga use case—an area where the scale and intricacy of constraints make conventional methods difficult to apply effectively.

Scientific Contribution and Innovation

From a scientific perspective, their contribution demonstrates the potential of automated heuristic discovery and adaptive planning algorithms. By combining AI program synthesis with the capabilities of large language models (LLMs), their approach is able to generate problem-specific strategies directly as code, moving beyond handcrafted heuristics. This enables scalable solutions that can adapt to evolving operational environments and tackle challenges at a level of complexity that is rarely manageable with conventional techniques.

This achievement also reflects the strong expertise of Hiverge in developing AI-driven planning technologies, building on their experience at the intersection of LLM-guided synthesis, reinforcement learning, and large-scale optimization.

The results of the Beluga AI Challenge confirm that methods capable of addressing both deterministic and probabilistic scalability are critical for the future of AI in planning domains—not only in aerospace logistics, but also in broader areas.

Scientific competitions are increasingly proving to be effective bridges between academia and industry. They push professionals to look at their challenges with fresh perspectives while giving students a realistic view of how problems are tackled in practice.

The TUPLES Beluga Challenge was so compelling that Linköping University launched a specialized PhD course: “Sequential Decision-Making at Airbus – The Beluga Challenge.”

Organized by the Machine Reasoning Lab, the course immersed students in an industrial AI application, guiding them to understand the problem in depth, design solutions, and exchange feedback. 

A team of ten PhD students and two professors from LiU submitted their work to the competition and achieved remarkable success: first place in the Explainability track.

A Comprehensive Approach to Complex AI Problems

The TUPLES Beluga Challenge models the logistics of transporting aircraft parts using Airbus Beluga airplanes. The core of the problem lies in managing the delivery of parts on “jigs” which need to be stored and reordered in capacity-limited racks, a task made difficult by potential congestion and the large number of jigs.
The LiU team tackled this intricate challenge head-on by creating three distinct systems:

• BeLiUga-Plan for the Scalability Deterministic track, where the team secured the third-place finish. This system, designed for scenarios with fixed flight schedules, addresses the challenge by decomposing the overall problem into a sequence of smaller, more manageable subtasks, one for each aircraft. This idea made it possible to find solutions for very large problem instances using off-the-shelf planning systems.
• BeLiUga-Reinforce for the Scalability Deterministic and Probabilistic tracks, the latter introducing the added complexity of uncertain flight arrival times. This solution utilized Proximal Policy Optimization (PPO), a reinforcement learning framework, and incorporated techniques like dynamic action masking to reduce the number of invalid actions and curriculum learning to progressively expose the agent to more difficult problems.
• BeLiUga-Explain for the Explainability track, a competition category the team won. This system was designed to answer a user’s questions about a given plan, thereby explaining the reasoning behind a planning system’s decisions. The system works by generating “counterfactuals”, which are alternative plans that adhere to the constraints of a specific question (e.g., “Why was jig X loaded on rack A instead of rack B?”) and then using a large language model to compare the original plan to the counterfactual to generate a clear, natural language explanation.

A Glimpse into the Student Experience

The course provided students with a unique opportunity to apply their academic knowledge to a challenging industrial application. The students describe the experience as a rewarding step from theoretical research to practical, real-world problem-solving.

“Working on a problem so closely related to an actual industrial application was incredibly eye-opening”, said Elliot Gestrin. “It really showed us the kind of challenges that exist beyond the academic benchmarks we’re used to.”

Paul Höft remarked, “The Beluga Challenge pushed us to think creatively and collaboratively. The problem was complex, and each track required a completely different approach. It was a true team effort to find solutions that were not only effective but also innovative.”

Mauricio Salerno, a guest PhD student from Spain, highlighted the significance of the Explainability track win: “Winning the Explainability track was particularly special. In a world where AI is becoming more and more integrated into critical systems, the ability to explain ‘why’ a decision was made is just as important as the decision itself. Our work shows that we’re ready to build AI that is not only smart but also trustworthy.”

The LiU team’s success highlights the value of the course in bridging the gap between academic research and industrial application, preparing the next generation of AI experts to tackle some of the world’s most complex problems.

The team lead by Daniel Gnad:  Elliot Gestrin, Arash Haratian, Paul Höft, Oliver Joergensen,  Arnaud Lequen, Mauricio Salerno, Jendrik Seipp, Gustaf Söderholm and Amath Sow.

Arash Haratian, Arnaud Lequen, Amath Sow and Oliver Joergensen participated in the Beluga Challenge but not in the Explainability Challenge.

This is the second article dedicated to the Beluga Challenge awardees.

Léna Aix, Benoît Bompol, and Jean Jodeau are members of the Research and Development team at SIA AI, part of Sia Partners, an international consulting firm founded 26 years ago and recognized for its expertise in strategy, digital transformation, and AI-driven solutions.

The theme of the competition — solving a real, high-stakes industrial logistics problem — motivated the team to take part. Their commitment was intense and their approach rigorous. While their solution was not among the official winners, the TUPLES judging committee decided to assign a Special Prize to acknowledge both the originality of their contribution and the remarkable effort behind it.

Tackling the Challenge

The Beluga Challenge was centered on a logistics planning problem proposed by Airbus: the storage and management of cargo transported by Beluga XL aircraft. Participants faced two scalability challenges:

Deterministic challenge: designing a sequence of actions ensuring that (1) all incoming parts are unloaded and stored in the correct order, (2) all required parts are delivered to production in sequence, and (3) all empty jigs are prepared for outgoing Beluga flights.

Probabilistic challenge: addressing uncertainty in flight arrivals, which alters the availability of jigs. Here, the goal was to design a policy (a state-to-action mapping) that performs reliably under variable conditions.

To foster real-world adoption, the competition also included an Explainability Challenge, asking participants to make solutions interpretable — a key requirement for building trust among expert human planners.

SIA AI’s Approach

The SIA AI team addressed this highly combinatorial problem through pure Operations Research methods, developing a solution that combined elegance and technical depth. Their heuristic construction algorithm integrated four main features:

🔹 A taboo list to avoid undoing recent actions
🔹 Customized scores to prioritize the most promising actions
🔹 A rollback mechanism to escape potential dead ends
🔹 A moving horizon method to anticipate future events

This method enabled them to deliver high-quality solutions within the constraints of the challenge, highlighting both ingenuity and adaptability.

A Broader Commitment

SIA AI counts more than 300 AI experts worldwide, supported by a dedicated team of 30 Operations Research specialists. This group is deeply committed to tackling optimization challenges across industries, working closely with engineering teams to address problems along the entire value chain.

Their participation in the Beluga Challenge, and the recognition they received, reflects the team’s ability to apply research-driven methods to highly complex industrial problems — an approach that embodies the spirit of innovation at the core of TUPLES.

We present an in-depth article on the paper “Modeling and Explaining an Industrial Workforce Allocation and Scheduling Problem” by Ignace Bleukx, Ryma Boumazouza, Tias Guns, Nadine Laage, and Guillaume Poveda, which received the Best Application Paper Award at the 31st International Conference on Principles and Practice of Constraint Programming (CP 2025).

Among the authors, Ryma Boumazouza, Ignace Bleukx, Tias Guns, and Guillaume Poveda are members of the TUPLES consortium, highlighting the project’s contribution to advancing real-world, trustworthy AI for planning and scheduling.

The Challenge: Planning at Scale

Aircraft manufacturing requires hundreds of logistical operations every day, from transporting parts and equipment to loading and unloading cargo.
This activity presents several challenges such as the scale of the problems, the need for fair workload distribution, and the need for methods for mitigating unforeseen disruptions due to technical malfunctions or incompatible weather conditions.
These tasks must be allocated to specialized teams with the right training and equipment, under strict deadlines.

Traditionally, teams of operational planners schedule and allocate tasks manually. This is not only time-consuming, but also vulnerable to disruptions such as:

* Delays in part deliveries
* Technical malfunctions
* Adverse weather conditions
* Unavailability of trained workers

In such cases, planners are forced to revise schedules under extreme time pressure, often dropping tasks or pushing them into the next planning horizon.

The Research: Smarter Models, Fairer Schedules

The paper introduces novel CP models that balance efficiency and fairness in workforce allocation.
Key contributions include:

• Workload Balancing: ensuring a fair distribution of tasks across teams, avoiding overburdening.
• Scalability: solving large instances with hundreds of daily tasks in reasonable time.
• Disruption Handling: using explainable AI methods to detect, explain, and repair infeasible schedules.

Explainability in Action

One of the most innovative aspects of the paper lies in its focus on explainability. When schedules are disrupted, understanding why a plan becomes infeasible is crucial for human operators.

The researchers applied techniques such as:
• Minimal Unsatisfiable Subsets (MUSes): identifying the exact set of constraints that cause infeasibility.
• Minimal Correction Subsets (MCSes): suggesting constraint relaxations that restore feasibility.

This dual approach not only provides planners with insights into the source of disruptions but also offers automated alternatives for repairing the schedule. In practice, this helps planners make faster, better-informed decisions during operational crises.

Impact and Future Directions

The award-winning research demonstrates the feasibility of deploying interactive decision-support systems in real industrial settings.

Such systems would allow planners to:
1. Automatically generate fair, optimized schedules.
2. Understand the causes of disruptions when plans fail.
3. Quickly review and adapt automated alternatives to restore feasibility.

Future work includes:
• Extending the problem from allocation and scheduling to full allocation-and-routing.
• Conducting user studies to evaluate the effectiveness of generated explanations in real planning environments.
• Investigating robust scheduling methods that anticipate disruptions before they occur.

🏆 Why It Matters

This award-winning paper exemplifies the mission of TUPLES: developing trustworthy, robust, and explainable AI systems for planning and scheduling. By combining advanced CP modeling with explainability techniques, the research paves the way for more resilient and human-centered decision support in critical industrial domains.

📖 Read the full paper here

The TUPLES Consortium is proud to announce the release of the final version of its Self-Assessment Tool (SAT) — a practical, easy-to-use online diagnostic survey designed to help developers and organizations evaluate the trustworthiness of Artificial Intelligence (AI) systems for planning and scheduling in work environments.

👉 Access the tool here: TUPLES Self-Assessment Tool

A Practical Approach to Trustworthy AI

AI is increasingly shaping decision-making in industries worldwide, from manufacturing and logistics to energy and transport. Yet, with its benefits come significant challenges: ensuring these systems are robust, transparent, safe, and aligned with ethical principles.

The TUPLES SAT was developed precisely to address this need. Building on EU Ethics Guidelines for Trustworthy AI and the experience gained throughout the TUPLES project, the SAT translates high-level principles into clear, actionable questions and recommendations tailored to the planning and scheduling domain.

Why the TUPLES SAT Stands Out

When the TUPLES team surveyed existing tools — including the well-known ALTAI framework — they identified key shortcomings:
• Too broad or too technical for varied AI use cases
• Lengthy questionnaires with heavy explanatory text
• Limited focus on technical aspects like data management and cybersecurity in relevant contexts
• Complex jargon that hindered usability for non-technical decision-makers

The SAT was designed to overcome these limitations, offering:

• A concise and targeted questionnaire
• Clear, actionable recommendations
• Focus on core technical and operational aspects
• An accessible format suitable for both technical experts and managerial staff

From Competition to Real-World Use

Initially tested by participants in the TUPLES Competition, the SAT benefited from real feedback from both researchers and practitioners. This input refined its usability, clarity, and relevance.
Now in its final version (released at Month 34 of the project), the SAT is ready for wider adoption across research institutions, industry, and public organizations engaged in planning and scheduling AI solutions.

Part of the Long-Term TUPLES Vision

The SAT is more than just a project deliverable — it is intended as a lasting resource for the AI community. By making trustworthy AI measurable and actionable, the SAT aims to foster systems that are not only effective in their purpose but also aligned with ethical standards, compliant with regulations, and designed for safe, transparent human-AI collaboration.

Whether you are developing AI-driven planning tools, evaluating third-party systems, or simply aiming to align with EU best practices, the TUPLES SAT will guide you step-by-step through a meaningful evaluation process.

In this article, we teamed up with Daniel Gnad and his brilliant team at Linköping University, who took home the Explainability Prize for teaching AI how to explain its own decisions.
Besides Daniel Gnad, the team consisted of his colleague Jendrik Seipp and four doctoral students: Elliot Gestrin, Gustaf Söderholm, Paul Höft, and Mauricio Salerno.

The Problem

Imagine you are managing a massive aircraft factory where giant cargo planes called Belugas arrive daily, delivering wings, turbines and other airplane parts. These components must be carefully stored on limited rack space before they can be assembled
into new aircraft.
Perhaps you would like an artificial intelligence system to solve this complex logistics problem, deciding where to place each part and when to move them around.
But, what happens when a human supervisor questions the AI’s choices?

“Why did you put that wing component on rack 3 instead of rack 7?”

To gain human trust and enable effective collaboration, it is crucial that AI systems can answer these kinds of questions. As such, TUPLES and Airbus organized the international Beluga AI Challenge, where one of the tasks was to develop an AI system capable of explaining itself.

The Swedish Solution

A team from Linköping University took part and won the competition with their system BeLiUga-Explain. Their breakthrough came from an approach called “counterfactual reasoning”, essentially teaching the AI to consider “what if” scenarios. When someone
questions a decision, the system creates an alternative version of the problem where that specific choice is forbidden or enforced. It then finds the best possible solution under these new constraints and compares it with the original.
For example, if asked why a wing component was stored on rack A instead of rack B, BeLiUga-Explain re-solves the entire logistics task with the restriction that the wing must go on rack B. It then compares the two solutions and uses modern large language models (LLMs) to explain the trade-offs in a human-understandable way:

“Using rack B would have required three additional component moves and thereby delayed assembly.”

More Than Just Moving Parts

The implications extend far beyond aircraft manufacturing. Daniel Gnad, the team leader and assistant professor at Linköping University, explains:

“This is not just about logistics – it is about trust. When AI systems make decisions that affect safety, efficiency, or costs, humans need to understand the reasoning behind those choices.”

The team consisted of several PhD students and two professors at Linköping University who, as part of a PhD course, combined their expertise in AI planning and human-computer interaction to design their solution. What makes this success remarkable is that the team was not specialized in explainable AI before. It was the unique combination of different expertise areas that made the difference.

Flying Onwards

Members of the team have since continued working on explainability solutions. Their newest system, called PlanPilot, allows users to interactively explore the diverse set of plans found by AI planning systems, through similarly enforcing or forbidding certain
choices, without needing to re-solve a modified problem.  Making systems capable of naturally and quickly interacting with humans like this is a key requirement for AI to be applicable in real-world scenarios.
The team will keep on working on their findings to make AI a transparent and trustworthy partner in complex decision-making.

Here’s a short overview of each contribution, with links to the corresponding papers for those who’d like to explore further.

Award-Winning Industrial Application

A major highlight was the paper “Modeling and Explaining an Industrial Workforce Allocation and Scheduling Problem” by Ignace Bleukx, Ryma Boumazouza, Tias Guns, Nadine Laage, and Guillaume Poveda, which received the Best Application Paper Award. The research addresses the complex task of allocating and scheduling workforces in aircraft manufacturing, a domain where disruptions due to technical failures, weather conditions, or resource unavailability are frequent.
The paper introduces advanced CP models that balance efficiency and fairness in workforce allocation, while also embedding explainability into the decision-support process. Techniques such as Minimal Unsatisfiable Subsets (MUSes) and Minimal Correction Subsets (MCSes) allow planners to understand why schedules fail and explore automated repair strategies. This award-winning work exemplifies TUPLES’ mission: developing trustworthy, robust, and explainable AI tools that can be deployed in real-world industrial contexts.
See the full paper here: https://bit.ly/41lVrpf

New Solver Architectures for Scheduling

Another paper, “Disjunctive Scheduling in Tempo” by Emmanuel Hebrard, presents Tempo, a novel hybrid CP/SAT solver. Tempo integrates lazy clause generation with difference logic, making it particularly effective for temporal and scheduling problems. Its design leverages edges in temporal graphs as branching variables, combining propagation and inference in an elegant and efficient way.
Experiments on job-shop scheduling benchmarks demonstrate that Tempo achieves competitive performance with state-of-the-art solvers like OR-Tools and CP Optimizer, while maintaining a simpler and more transparent architecture. This work underscores the importance of solver design choices and opens promising directions for extending hybrid methods to broader classes of scheduling and resource allocation problems.

See the full paper here:https://bit.ly/45KmyLX

Understanding the Role of Heuristics

The third contribution, “Understanding the Impact of Value Selection Heuristics in Scheduling Problems” by Tim Luchterhand, Emmanuel Hebrard, and Sylvie Thiébaux, sheds light on the often-overlooked role of value selection heuristics in constraint programming.
Through extensive experiments, the authors show that while value heuristics generally have less impact than variable ordering, their role becomes critical in hard scheduling problems. Surprisingly, they find that heuristic accuracy tends to decrease as the problem becomes tighter, exactly when mistakes become most costly. Moreover, sophisticated heuristics, while more accurate overall, pay a disproportionately higher price when they make an error.
This nuanced analysis provides a deeper understanding of why designing impactful value selection heuristics is challenging, offering guidance for future heuristic design and integration with machine learning approaches.

See the full paper here: https://bit.ly/4oVrN4v

Prominent Paper Award

We are also delighted to celebrate Hélène Verhaeghe, who was a member of the TUPLES consortium, for receiving the Prominent Paper Award of the Constraints Journal for her work “Learning Optimal Decision Trees Using Constraint Programming”, co-authored with Siegfried Nijssen, Gilles Pesant, Claude-Guy Quimper, and Pierre Schaus.

A Strong Presence at CP 2025

Together, these three papers illustrate the breadth and depth of TUPLES’ contributions to the constraint programming community: from award-winning industrial applications, to innovative solver architectures, to fundamental insights into search heuristics.
By combining applied impact with theoretical advances, TUPLES researchers are pushing the boundaries of trustworthy, explainable, and high-performance AI for scheduling—advancing both the science and its deployment in real-world domains.