DA-SLAM: Deep Active SLAM based on Deep Reinforcement Learning

  • Martin Alcalde Facultad de Ingeniería
  • Matias Ferreira Facultad de Ingeniería
  • Pablo González Facultad de Ingeniería
  • Federico Andrade Facultad de Ingeniería
  • Gonzalo Tejera Facultad de Ingeniería


This work presents improvements to the state-of-the-art algorithms for path planning and exploration of unknown and complex environments using Deep Reinforcement Learning. Our novel approach takes into consideration: (i) map information, built online by the robot using a Simultaneous Localization and Mapping algorithm and (ii) uncertainty of the robot's pose, which leads to active loop-closing to encourage exploration and better map generation within two agents. The results show that the map completeness-based reward function outperforms literature's results on shorter trajectories, thus, better performance; while uncertainty-based with loop-closing reward function improves map generation. Both agents showed the ability, to perform Active SLAM over complex environments and generalization to unseen maps capabilities.
Palavras-chave: Deep learning, Simultaneous localization and mapping, Uncertainty, Computational modeling, Education, Reinforcement learning, Entropy, SLAM, Active SLAM, AI, Machine Learning, Computer Vision, Robotics, RL, DRL, PPO, Open AI, ROS
Como Citar

Selecione um Formato
ALCALDE, Martin; FERREIRA, Matias; GONZÁLEZ, Pablo; ANDRADE, Federico; TEJERA, Gonzalo. DA-SLAM: Deep Active SLAM based on Deep Reinforcement Learning. In: SIMPÓSIO BRASILEIRO DE ROBÓTICA E SIMPÓSIO LATINO AMERICANO DE ROBÓTICA (SBR/LARS), 19. , 2022, São Bernardo do Campo/SP. Anais [...]. Porto Alegre: Sociedade Brasileira de Computação, 2022 . p. 282-287.