• Lucas Pereira Cotrim ([email protected])
• Marcos Menon José ([email protected])
• Eduardo Lobo Lustosa Cabral ([email protected])

Reinforcement Learning Framework for simulation-based positioning control of various robotic manipulators. Two main RL algorithms are implemented: REINFORCE and DQN. For each project, the main file is "<project_name>.m", for example: DQNKuka.m.

For more information please refer to the article "Reinforcement Learning Control of Robot Manipulator" and to the bachelor thesis "Controle de Robô Manipulador com Aprendizado por Reforço", currently available in portuguese.

• PGRobotArmR: 1-DoF robot.

• PGRobotArmRR: 2-DoF robot.

• PGKuka: Kuka robot (5 controllable DoF).

• PGKukaUnknownObstacle: Kuka robot in environment with unknown obstacle (Wall).

• PGKukaRandomPositions: Kuka robot trained for generic goal and obstacle positions.

• QLearningRobotArmR: 1-DoF robot.

• QLearningRobotArmRR: 2-DoF robot.

• DQNKuka: Kuka robot (5 controllable DoF).

• DQNKukaUnknownObstacle: Kuka robot in environment with unknown obstacle (Wall).

• DQNKukaRandomPositions: Kuka robot trained for generic goal and obstacle positions.

• DQNKukaImage: Kuka robot DQN agent trained with image-based states.