Sample-Efficient Learning to Solve a Real-World Labyrinth Game Using Data-Augmented Model-Based Reinforcement Learning

Motivated by the challenge of achieving rapid learning in physical environments, this paper presents the development and training of a robotic system designed to navigate and solve a labyrinth game using model-based reinforcement learning techniques. The method involves extracting low-dimensional observations from camera images, along with a cropped and rectified image patch centered on the current position within the labyrinth, providing valuable information about the labyrinth layout. The learning of a control policy is performed purely on the physical system using model-based reinforcement learning, where the progress along the labyrinth’s path serves as a reward signal. Additionally, we exploit the system’s inherent symmetries to augment the training data. Consequently, our approach learns to successfully solve a popular real-world labyrinth game in record time, with only 5 hours of real-world training data.

Authors

• Thomas Bi
• Raffaello D'Andrea

Keywords

• Training
• Learning systems
• Navigation
• Robot vision systems
• Training data
• Games
• Hardware
• Model-based Reinforcement Learning
• Sample-efficient Learning
• Labyrinth Game
• Current Position
• Physical System
• Camera Images
• Recording Time
• Image Patches
• Policy Learning
• Popular Game
• Reinforcement Learning Techniques
• Time Step
• Active Control
• Data Augmentation
• Visuospatial
• Angular Velocity
• Rotation Axis
• Inclination Angle
• Deep Reinforcement Learning
• Ball Position
• Model-based Control
• Physical Setup
• Direction Of Path
• Wide-angle Camera
• Replay Buffer