MarkFzp/mobile-aloha
Mobile Aloha
Mobile Alohas هو إطار عمل للتلاعب المحمول ثنائي اليد مصمم لتعلم التحكم في الروبوت لكامل الجسم. يوفر خط أنابيب متكامل لتعلم التقليد يدير عملية جمع بيانات العرض البشري وتدريب نماذج السلوك لأتمتة المهام الجسدية المعقدة.
يتميز النظام بواجهة تشغيل روبوت عن بعد تربط حركات الإنسان بروبوت محمول بذراعين. يتضمن أداة مجموعة بيانات حركة كامل الجسم تُستخدم لتسجيل وتصور وإعادة تشغيل بيانات المفاصل والمستشعرات من جلسات التلاعب.
يغطي إطار العمل العديد من مجالات القدرة، بما في ذلك التشغيل عن بعد لكامل الجسم، وتدريب تعلم التقليد، والتلاعب الروبوتي ثنائي اليد. كما يوفر أدوات لتصور البيانات المسجلة وإعادة تشغيل حركة الأجهزة.
Features
- Dual-Arm Mobile Robots - Provides a framework for controlling a robotic platform combining a wheeled base with two arms.
- Behavioral Cloning Toolkits - Trains behavior models using human demonstration data to automate complex physical movements.
- Imitation Learning Pipelines - Implements a workflow for collecting human demonstration data and training behavior models.
- Policy Execution Engines - Runs trained behavior models to autonomously generate control commands based on live sensory input.
- Imitation Learning Trainings - Trains neural networks to map sensor inputs to robot actions using human demonstration data.
- Teleoperated Demonstration Collection - Captures high-dimensional movement data by mapping human controllers to a remote robot in real time.
- Bimanual Mobile Manipulation Frameworks - Provides a complete framework for learning whole-body robot control through human teleoperation and imitation learning.
- Teleoperation Systems - Ships a remote control system that maps human movements to a mobile robot with dual arms.
- Demonstration Collection Systems - Captures and structures sensor and joint data from human operators for machine learning datasets.
- Bimanual Coordination - Implements unified control and learning for synchronized dual-arm and mobile base movements.
- Whole-Body Teleoperation - Enables real-time remote control of the mobile robot's entire body and arms using a master device.
- Bimanual Capture - Records human demonstrations of dual-arm and base movements for bimanual robotic learning.
- Mobile Base Capture - Captures whole-body human movements via a remote interface to create imitation learning datasets.
- Robotic Motion Dataset Tools - Provides a utility for recording, visualizing and replaying joint and sensor data from manipulation sessions.
- Hardware Movement Replays - A functionality that executes previously saved episodes by sending stored movement commands back to the physical hardware.
- Robotics Visualization Tools - Provides tools for reviewing and rendering recorded sensor data and physical episodes.
- Robotic State Logging - Saves timestamped robotic joint angles and sensor readings into structured files for dataset generation.
- Data Stream Recording and Replay - Provides capabilities to stream saved joint commands back to physical hardware for episode replay.
- Robotics Data Recording - Captures and saves joint and sensor information during sessions into structured files.
