Constrained Behavior Cloning for Robotic Learning

Behavior cloning (BC) is a widely used method for learning from expert demonstrations due to its simplicity and efficiency. However, the reliability and stability of BC decline when facing data distribution shifts, especially in single-arm robots with limited fields of view. This study introduces a Geometrically and Historically Constrained Behavior Cloning (GHCBC) method, where an HCBC module utilizes visual and action histories to capture temporal dependencies, maximizing the use of available information, and a GCBC module incorporates high-level perceptual data, such as the relative poses of joints and end-effectors, to enhance BC performance. Experiments demonstrate that the GHCBC outperforms current SOTA BC methods, achieving a 31. 5% improvement in simulation success rates and 48. 4% in real-robot scenarios respectively. To the best of our knowledge, this is the first time that the GHCBC has been introduced in robotic BC where great potential is demonstrated for long-term tasks in real world environments.

Authors

• Jun Xie
• Jianwei Tan
• Wensheng Liang
• Zhicheng Wang
• Xiaoguang Ma

Keywords

• Visualization
• Dynamics
• Cloning
• Transformers
• End effectors
• Vectors
• Stability analysis
• History
• Reliability
• Intelligent robots
• Behavior Cloning
• Domain Shift
• Real-world Environments
• Relative Pose
• Limited Field Of View
• Use Of Available Information
• Expert Demonstrations
• Transformer
• Dynamic Environment
• Visual Features
• Simulation Environment
• Historical Information
• Red Box
• Robotic Arm
• Geometric Constraints
• Single Camera
• Real Robot
• Imitation Learning
• Positional Encoding
• End-effector Pose
• Average Success Rate
• Color Block
• Pose Information
• Current Pose
• Policy Execution
• Information Compression
• Human Learning
• Task Execution
• History Of Activity