Juan Ignacio Vazquez
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Juan Ignacio Vazquez has a total of 63 co-authors. Here are the top co-authors by number of shared publications:
- Diego López-De-Ipiña
University of Deusto, Bilbao, Spain
13 shared publications - Aitor Almeida
DeustoTech, University of Deusto, Bilbao, Spain
5 shared publications - Asier Gonzalez-Santocildes
Faculty of Engineering, University of Deusto, Bilbao, Spain
5 shared publications - Andoni Eguiluz
University of Deusto
4 shared publications - Alberto Tellaeche
University of Deusto
3 shared publications - Ignacio Fidalgo Astorquia
Department of Computing, Electronics and Communication Technologies, University of Deusto, Avenida de las Universidades 24, 48007 Bilbao, Spain
3 shared publications
Latest Publications
- Open Access
- 17 October 2026
- Published by Elsevier in SSRN Electronic Journal
Comfort-Aware Reinforcement Learning for Adaptive Trajectory Planning in Human--Robot Collaboration
Ensuring human comfort during physical human–robot interaction remains a fundamental challenge in collaborative robotics. This study presents a real-world evaluation of reinforcement learning controllers trained with comfort-aware reward functions for trajectory planning in a UR3e manipulator. Fifty participants, stratified by age, gender, and prior knowledge of robotics, experienced three robot trajectories while physiological signals (heart rate changes and movement indices) and subjective ratings were recorded. Unsupervised clustering revealed three distinct physiological reactivity profiles that aligned with both objective and self-reported comfort measures. These profiles guided real-time adaptation of robot motion, enabling trajectory personalization based on individual physiological responses. A simple distance-to-centroid classifier was derived to assign new users to comfort profiles, allowing immediate deployment without continuous sensing. These findings provide empirical support for integrating human-centered reward shaping into reinforcement-learning-based control policies and offer a scalable pathway toward personalized, comfort-optimized robot behavior in real-world settings.
- 17 October 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessComfort-Aware Reinforcement Learning for Adaptive Trajectory Planning in Human--Robot Collaboration
Ensuring human comfort during physical human–robot interaction remains a fundamental challenge in collaborative robotics. This study presents a real-world evaluation of reinforcement learning controllers trained with comfort-aware reward functions for trajectory planning in a UR3e manipulator. Fifty participants, stratified by age, gender, and prior knowledge of robotics, experienced three robot trajectories while physiological signals (heart rate changes and movement indices) and subjective ratings were recorded. Unsupervised clustering revealed three distinct physiological reactivity profiles that aligned with both objective and self-reported comfort measures. These profiles guided real-time adaptation of robot motion, enabling trajectory personalization based on individual physiological responses. A simple distance-to-centroid classifier was derived to assign new users to comfort profiles, allowing immediate deployment without continuous sensing. These findings provide empirical support for integrating human-centered reward shaping into reinforcement-learning-based control policies and offer a scalable pathway toward personalized, comfort-optimized robot behavior in real-world settings.
- Open Access
RehabBot: Reinforcement learning environment for upper and lower limb robot-assisted rehabilitation
- Open Access
Dissecting Quantum Reinforcement Learning: A Systematic Evaluation of Key Components
Parameterised quantum circuit (PQC) based Quantum Reinforcement Learning (QRL) has emerged as a promising paradigm at the intersection of quantum computing and reinforcement learning (RL). By design, PQCs create hybrid quantum-classical models, but their practical applicability remains uncertain due to training instabilities, barren plateaus (BPs), and the difficulty of isolating the contribution of individual pipeline components. In this work, we dissect PQC based QRL architectures through a systematic experimental evaluation of three aspects recurrently identified as critical: (i) data embedding strategies, with Data Reuploading (DR) as an advanced approach; (ii) ansatz design, particularly the role of entanglement; and (iii) post-processing blocks after quantum measurement, with a focus on the underexplored Output Reuse (OR) technique. Using a unified PPO-CartPole framework, we perform controlled comparisons between hybrid and classical agents under identical conditions. Our results show that OR, though purely classical, exhibits distinct behaviour in hybrid pipelines, that DR improves trainability and stability, and that stronger entanglement can degrade optimisation, offsetting classical gains. Together, these findings provide controlled empirical evidence of the interplay between quantum and classical contributions, and establish a reproducible framework for systematic benchmarking and component-wise analysis in QRL.
Open AccessDissecting Quantum Reinforcement Learning: A Systematic Evaluation of Key Components
Parameterised quantum circuit (PQC) based Quantum Reinforcement Learning (QRL) has emerged as a promising paradigm at the intersection of quantum computing and reinforcement learning (RL). By design, PQCs create hybrid quantum-classical models, but their practical applicability remains uncertain due to training instabilities, barren plateaus (BPs), and the difficulty of isolating the contribution of individual pipeline components. In this work, we dissect PQC based QRL architectures through a systematic experimental evaluation of three aspects recurrently identified as critical: (i) data embedding strategies, with Data Reuploading (DR) as an advanced approach; (ii) ansatz design, particularly the role of entanglement; and (iii) post-processing blocks after quantum measurement, with a focus on the underexplored Output Reuse (OR) technique. Using a unified PPO-CartPole framework, we perform controlled comparisons between hybrid and classical agents under identical conditions. Our results show that OR, though purely classical, exhibits distinct behaviour in hybrid pipelines, that DR improves trainability and stability, and that stronger entanglement can degrade optimisation, offsetting classical gains. Together, these findings provide controlled empirical evidence of the interplay between quantum and classical contributions, and establish a reproducible framework for systematic benchmarking and component-wise analysis in QRL.
- 14 October 2025
- Published by Springer Nature
- p. 251-262
Enhancing World Models with Specialized Prediction Networks for Reinforcement Learning
- 14 October 2025
- Published by Springer Nature
- p. 251-262
Enhancing World Models with Specialized Prediction Networks for Reinforcement Learning
- 14 October 2025
- Published by Springer Nature
- p. 239-250
SimplyQRL: A Modular Benchmarking Library for Hybrid Quantum Reinforcement Learning
- 14 October 2025
- Published by Springer Nature
- p. 239-250
SimplyQRL: A Modular Benchmarking Library for Hybrid Quantum Reinforcement Learning
- Open Access
Comparative Benchmark of Sampling-Based and DRL Motion Planning Methods for Industrial Robotic Arms
This study presents a comprehensive comparison between classical sampling-based motion planners from the Open Motion Planning Library (OMPL) and a learning-based planner based on Soft Actor–Critic (SAC) for motion planning in industrial robotic arms. Using a UR3e robot equipped with an RG2 gripper, we constructed a large-scale dataset of over 100,000 collision-free trajectories generated with MoveIt-integrated OMPL planners. These trajectories were used to train a DRL agent via curriculum learning and expert demonstrations. Both approaches were evaluated on key metrics such as planning time, success rate, and trajectory smoothness. Results show that the DRL-based planner achieves higher success rates and significantly lower planning times, producing more compact and deterministic trajectories. Time-optimal parameterization using TOPPRA ensured the dynamic feasibility of all trajectories. While classical planners retain advantages in zero-shot adaptability and environmental generality, our findings highlight the potential of DRL for real-time and high-throughput motion planning in industrial contexts. This work provides practical insights into the trade-offs between traditional and learning-based planning paradigms, paving the way for hybrid architectures that combine their strengths.
Open AccessComparative Benchmark of Sampling-Based and DRL Motion Planning Methods for Industrial Robotic Arms
This study presents a comprehensive comparison between classical sampling-based motion planners from the Open Motion Planning Library (OMPL) and a learning-based planner based on Soft Actor–Critic (SAC) for motion planning in industrial robotic arms. Using a UR3e robot equipped with an RG2 gripper, we constructed a large-scale dataset of over 100,000 collision-free trajectories generated with MoveIt-integrated OMPL planners. These trajectories were used to train a DRL agent via curriculum learning and expert demonstrations. Both approaches were evaluated on key metrics such as planning time, success rate, and trajectory smoothness. Results show that the DRL-based planner achieves higher success rates and significantly lower planning times, producing more compact and deterministic trajectories. Time-optimal parameterization using TOPPRA ensured the dynamic feasibility of all trajectories. While classical planners retain advantages in zero-shot adaptability and environmental generality, our findings highlight the potential of DRL for real-time and high-throughput motion planning in industrial contexts. This work provides practical insights into the trade-offs between traditional and learning-based planning paradigms, paving the way for hybrid architectures that combine their strengths.
- 9 October 2024
- Published by Springer Nature
- p. 239-251
Reinforcement Learning for Dynamic Trajectory Adjustment in Human-Robot Interaction Within Virtual Simulations
- 9 October 2024
- Published by Springer Nature
- p. 239-251
Reinforcement Learning for Dynamic Trajectory Adjustment in Human-Robot Interaction Within Virtual Simulations
- 9 October 2024
- Published by Springer Nature
- p. 280-291
Unity-Based Autonomous Driving Environment: A Platform for Validating Reinforcement Learning Agents
- 9 October 2024
- Published by Springer Nature
- p. 280-291
Unity-Based Autonomous Driving Environment: A Platform for Validating Reinforcement Learning Agents
- 9 October 2024
- Published by Springer Nature
- p. 292-303
Using Federated Learning Techniques to Generalize Reinforcement Learning Approaches
- 9 October 2024
- Published by Springer Nature
- p. 292-303
Using Federated Learning Techniques to Generalize Reinforcement Learning Approaches
- Open Access
- 29 August 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 122289-122299
Adaptive Robot Behavior Based on Human Comfort Using Reinforcement Learning
This study explores the potential of training robots using reinforcement learning (RL) to adapt their behavior based on human comfort levels during tasks. An experimental environment has been developed and made available to the research community, facilitating the replication of these experiments. The results demonstrate that adjusting a single comfort-related input parameter during training leads to significant variations in the robot’s behavior. Detailed discussions of the reward functions and obtained results validate these behavioral adaptations, confirming that robots can dynamically respond to human needs, thereby enhancing human-robot interaction. While the study highlights the effectiveness of this approach, it also raises the question of real-time comfort measurement, suggesting various systems for future exploration. These findings contribute to the development of more intuitive and emotionally responsive robots, offering new possibilities for future research in advancing human-robot interaction.
- 29 August 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 122289-122299
Open AccessAdaptive Robot Behavior Based on Human Comfort Using Reinforcement Learning
This study explores the potential of training robots using reinforcement learning (RL) to adapt their behavior based on human comfort levels during tasks. An experimental environment has been developed and made available to the research community, facilitating the replication of these experiments. The results demonstrate that adjusting a single comfort-related input parameter during training leads to significant variations in the robot’s behavior. Detailed discussions of the reward functions and obtained results validate these behavioral adaptations, confirming that robots can dynamically respond to human needs, thereby enhancing human-robot interaction. While the study highlights the effectiveness of this approach, it also raises the question of real-time comfort measurement, suggesting various systems for future exploration. These findings contribute to the development of more intuitive and emotionally responsive robots, offering new possibilities for future research in advancing human-robot interaction.