Titulo Estágio
Learning from Human Demonstration for Real-Time Control in Stone-Cutting Operations
Áreas de especialidade
Sistemas Inteligentes
Local do Estágio
CISUC
Enquadramento
The stone processing industry — including the cutting and shaping of materials such as marble, granite, and engineered stone — is undergoing a transformative shift driven by Industry 4.0 technologies. While modern cutting machines now incorporate advanced sensor systems, CNC controllers, and automation interfaces, a large portion of critical decision-making during operation still depends on skilled human operators.
These operators handle complex and delicate tasks, such as:
• Positioning the cutting tool based on stone geometry, texture, and defects,
• Adjusting cutting parameters in real time depending on vibration, noise, tool wear, or material hardness, and
• Making safety-critical judgments under changing environmental and operational conditions.
Traditional automation using hard-coded logic (PLC) or fixed control algorithms (e.g., PID) cannot cope effectively with the variability in stone types, slab defects, custom orders, or tool dynamics. This often leads to inefficiencies, downtime, and inconsistent quality, particularly in short production batches or customized cuts.
Real-time control of cutting processes in the stone-processing industry demands constant monitoring and adjustment based on complex, multi-sensory feedback (e.g., vibration, force, acoustic signals). Despite the integration of advanced CNC machines and sensors, the core decision-making remains manual, relying on operator intuition to adapt cutting parameters to dynamic conditions. Traditional automation systems (e.g., PID controllers) lack the flexibility and learning capacity to handle such variability, and opaque AI systems pose challenges in terms of safety and user trust. There is a need for adaptive, explainable AI systems that can collaborate with human operators to perform real-time process control.
Objetivo
To develop a human-in-the-loop AI system that learns from operator behavior and sensor data to perform real-time control and adjustment of cutting parameters in response to dynamic conditions during stone processing.
The system will be trained using Learning from Demonstration (LfD) techniques and will operate in live or simulated environments using actual sensor inputs from the machinery.
Research Objectives:
1. Characterize real-time control tasks in stone cutting, including typical adjustments made by operators.
2. Identify relevant sensor inputs (e.g., force, vibration, acoustic feedback) and establish how they inform parameter tuning.
3. Design a learning framework to train an AI agent via demonstrations or feedback on dynamic control behaviors.
4. Integrate the AI agent into the control loop of a real or simulated stone-cutting system.
5. Assess performance in terms of responsiveness, cutting quality, and reduction in manual intervention.
Plano de Trabalhos - Semestre 1
1- State of the art [Sept – Oct]
2- Problem statement, research aims and objectives [Nov]
3- Design and first implementation of the AI system [Nov – Jan]
4- Thesis proposal writing [Dec – Jan]
Plano de Trabalhos - Semestre 2
5- Improvement of the AI system [Feb – Apr]
6- Experimental Tests [Apr – May]
7- Paper writing [May – Jun]
8- Thesis writing [Jan – Jul]
Condições
The work should take place at the Centre for Informatics and Systems of the University of Coimbra (CISUC) at the Department of Informatics Engineering of the University of Coimbra.
Observações
References
[1] 1. Argall, B. D., Chernova, S., Veloso, M., & Browning, B. (2009). A survey of robot learning from demonstration. Robotics and Autonomous Systems, 57(5), 469–483. https://doi.org/10.1016/j.robot.2008.10.024
[2] Christiano, P. F., Leike, J., Brown, T., Martic, M., Legg, S., & Amodei, D. (2017). Deep reinforcement learning from human preferences. Advances in Neural Information Processing Systems (NeurIPS), 30. https://arxiv.org/abs/1706.03741
[3] Reddy, S., Dragan, A., & Levine, S. (2019). Sqil: Imitation learning via regularized behavioral cloning. arXiv preprint arXiv:1905.11108. https://arxiv.org/abs/1905.11108
[4] Amershi, S., Cakmak, M., Knox, W. B., & Kulesza, T. (2014). Power to the people: The role of humans in interactive machine learning. AI Magazine, 35(4), 105–120. https://doi.org/10.1609/aimag.v35i4.2513
Orientador
Luís Miguel Machado Lopes Macedo
macedo@dei.uc.pt 📩