Boston Dynamics and Toyota Research Institute (TRI) have joined forces to revolutionize humanoid robotics by integrating TRI's advanced AI technology, specifically Large Behavior Models (LBMs), into Boston Dynamics' Atlas robot. This collaboration aims to create more adaptable and intelligent robots capable of learning complex tasks through observation and minimal demonstrations, potentially transforming industries from manufacturing to healthcare.
Boston Dynamics-TRI Partnership
The collaboration between Boston Dynamics and Toyota Research Institute (TRI) was announced on October 16, 2024, bringing together two industry leaders to push the boundaries of humanoid robotics. This partnership aims to integrate TRI's expertise in Large Behavior Models (LBMs) with Boston Dynamics' advanced Atlas robot, creating more adaptable and intelligent humanoid robots capable of learning and performing a wide range of tasks. The joint effort will be led by Scott Kuindersma from Boston Dynamics and Russ Tedrake from TRI, with research taking place in Boston.
Atlas Robot EnhancementsÂ
The latest electric version of Atlas, unveiled in April 2024, serves as the hardware platform for implementing TRI's advanced AI technologies. Key enhancements include:
Improved bimanual manipulation for a wide range of tasks
Integration of Large Behavior Models (LBMs) for rapid skill acquisition
Enhanced whole-body mobility and dexterity
Capability to learn from human demonstrations and adapt to new situations
Improved sensors and data collection systems for performance evaluation and AI model training
These advancements are expected to push Atlas beyond its current capabilities, potentially enabling it to perform complex tasks with greater autonomy and flexibility in various real-world scenarios.
AI Applications and ImpactsÂ
The integration of TRI's Large Behavior Models (LBMs) into Atlas enables rapid acquisition of new, robust, dexterous skills across the entire body, allowing the robot to tackle diverse challenges in dynamic environments. Potential applications extend beyond traditional industrial tasks to include elder care, household assistance, and emergency response. The partnership aims to implement vision-and-language-oriented foundational models for advanced manipulation tasks, drawing on TRI's research in computer vision and large language models. This collaboration is expected to accelerate the development of versatile humanoid robots, increase competition in the market, and potentially lead to faster commercialization of humanoid robots for both industrial and household applications. The focus on human-robot interaction and safety considerations is likely to become more prevalent as these technologies advance.
TRI's AI ApproachÂ
Large Behavior Models (LBMs) developed by TRI represent a significant departure from traditional AI approaches in robotics. These models enable robots to learn complex tasks through observation and minimal demonstrations, similar to how large language models function in natural language processing. TRI's innovative diffusion policy technique allows robots to generate and refine a range of possible actions over time, enhancing their ability to perform intricate maneuvers with fewer training examples.
Rapid learning: Tasks that previously took weeks can now be mastered in hours
Flexibility: Robots can adapt to various tasks and environments with minimal reprogramming
Efficiency: Reduces time and resources needed for training compared to conventional methods
Generalization: Ability to apply learned skills across different scenarios and objects
If you work within a business and need help with AI, then please email our friendly team via admin@aisultana.com .
Try the AiSultana Wine AI consumer application for free, please click the button to chat, see, and hear the wine world like never before.
Comentarios