DyLam
AAMAS 2025
Dynamic Lambda: a method for studying reward signals and their impact over time in multi-agent reinforcement learning environments.
Read PaperPublications
Planning the Path with Reinforcement Learning
arXiv, 2024
This work investigates the potential of Reinforcement Learning to tackle robot motion planning challenges in the dynamic RoboCup Small Size League. Using a heuristic control approach, we evaluate RL's effectiveness in obstacle-free and single-obstacle path-planning environments. Our method achieved a 60% time gain in obstacle-free environments compared to baseline algorithms, and demonstrated dynamic obstacle avoidance capabilities.
Read Paper
rSoccer: A Framework for Studying RL in Robot Soccer
arXiv, 2021
Introduces an open-source simulator for the IEEE Very Small Size Soccer and the Small Size League optimized for reinforcement learning experiments. Proposes a framework for creating OpenAI Gym environments with benchmark tasks for evaluating single-agent and multi-agent robot soccer skills.
Read Paper
RL Applied to Coach Task in IEEE Very Small Size Soccer
Latin American Robotics Symposium, 2020
Proposes an end-to-end approach for the coaching task based on Reinforcement Learning. The system processes information during simulated matches to learn an optimal policy that chooses the current formation depending on the opponent and game conditions. Achieved a win/loss ratio of approximately 2.0 against one of the top teams of the VSSS league.
Read Paper
RL-driven Automatic Calibration for Color Segmentation-based Robot Detection
IEEE, 2024
Applies reinforcement learning to automate the calibration process for color segmentation-based robot detection systems, improving accuracy and reducing manual tuning effort.
Read Paper
Segment Routing Path Optimization for URLLC via Multi-Armed Bandits
IEEE, 2025
Applies multi-armed bandit algorithms to optimize segment routing paths for Ultra-Reliable Low-Latency Communications in next-generation mobile transport networks.
Read Paper