How does artificial intelligence actually think? Why does it make a particular decision in a given situation? These are the questions addressed by PANSim, a tool developed as the bachelor’s thesis of CTU Faculty of Electrical Engineering student Bc. Erol Medenčević under the supervision of Ing. Jakub Med, in collaboration with the team of Assoc. Prof. Lukáš Chrpa from the Department of Industrial Informatics at the Czech Institute of Informatics, Robotics and Cybernetics, CTU (CIIRC CTU). The project received the Best Demonstration Award on 1 July 2026 at ICAPS 2026 in Dublin, Ireland – the world’s leading conference on automated planning.
As artificial intelligence becomes increasingly integrated into more areas of everyday life, the need to understand why it makes particular decisions is also growing. The award is particularly valuable because the winner was not selected by a pre-appointed jury but by the conference participants themselves – the global community of researchers working in automated planning and AI decision-making. Among twelve selected demonstration projects from leading universities and research institutions around the world, PANSim received the highest number of votes.
“This is an exceptionally valuable award for us because ICAPS is the world’s leading conference on automated planning. Unlike large conferences covering artificial intelligence in general, ICAPS brings together virtually the entire research community in our field. That makes it even more meaningful that the conference participants themselves selected our tool as the best demonstration of this year’s conference,” says Assoc. Prof. Lukáš Chrpa from the Department of Industrial Informatics at the Czech Institute of Informatics, Robotics and Cybernetics, CTU.
When It Is Not Enough to Know That AI Made the Right Decision
Artificial intelligence is now used to control robots, autonomous vehicles and industrial systems. However, understanding why an algorithm made a particular decision is often much more difficult than verifying whether the final result is correct. Even the developers themselves are not always able to understand complex decision-making processes simply by examining text logs or program outputs.
PANSim transforms AI decision-making into an interactive graphical simulation. It focuses on the field of symbolic artificial intelligence, which relies on explicit modelling of the environment and step-by-step planning rather than statistical learning. Users can observe, step by step, how an agent evaluates the situation, responds to changes in its environment, and searches for a safe path to its goal. This makes it much easier to identify algorithmic errors, understand the agent’s reasoning, and further improve the overall system.
“It is not just about the graphics themselves. The visual representation makes it much easier to understand what is actually happening during the decision-making process. Developers can quickly see why the agent behaved in a particular way and where the algorithm needs to be modified,” explains Ing. Jakub Med, supervisor of Erol Medenčević’s bachelor’s thesis and a PhD student at the CTU Faculty of Electrical Engineering, who also works at the Department of Industrial Informatics at CIIRC CTU.
How can the complex decision-making of artificial intelligence be presented as clearly as possible? The authors chose a simple scenario. A frog has to cross a pond and collect all the coins. However, it is not enough to find the shortest route, because some lily pads may disappear beneath the water at any moment. The agent therefore has to continuously reassess its options and respond to random changes in the environment. A second scenario simulates autonomous underwater vehicles that must safely react to the movement of ships above the water while carrying out their mission. The team also requires the agent to provide a guarantee of success, distinguishing the techniques developed by the team from traditional approaches based on machine learning algorithms.
From a Bachelor’s Thesis to Further Research
PANSim is not the first international success of CTU Faculty of Electrical Engineering student Erol Medenčević, who continues to develop the project during his master’s studies. Earlier this year, the project was presented at the prestigious AAAI Conference in Singapore. While AAAI is one of the world’s largest artificial intelligence conferences, attracting thousands of participants, ICAPS is the leading international forum dedicated specifically to automated planning.
“After the project was accepted to a top international conference, I realised that it had real impact. But when my colleagues sent me a photo of the Best Demonstration Award certificate from Dublin, I was genuinely surprised. It is a great motivation to continue developing the project,” says Bc. Erol Medenčević.
In the next phase of development, the authors plan to make the agents’ decision-making even more transparent. Planned features include the ability to rewind the simulation by several steps, better visualisation of the AI’s current intentions, and additional simulation scenarios. Their long-term ambition is to create a universal environment capable of visualising a broad range of planning tasks.
Although PANSim is currently intended primarily for researchers, the principles it employs have practical applications in areas such as autonomous robotics and cybersecurity, where systems must respond to unpredictable changes in their environment. As the number of autonomous systems continues to grow, a deeper understanding of how artificial intelligence reasons and why it makes particular decisions will become increasingly important.
