Boston Dynamics and Former CEO Team Up to Accelerate Atlas' Learning

Boston Dynamics, the global leader in developing highly mobile robots, has joined forces with its former CEO, Marc Raibert, to speed up the learning process of its Atlas humanoid robot. The collaboration aims to advance the capabilities of Atlas through reinforcement learning, making it more flexible and adaptable to various environments and applications.

The partnership between Boston Dynamics and the Robotics & AI Institute (formerly The AI Institute), co-founded by Raibert, brings together two of the world's leading research teams in robotics and AI. By combining their expertise, the organizations hope to accelerate the development of reinforcement learning for Atlas, bridging the sim-to-real gap, and improving whole-body locomotion and manipulation.

Atlas, Boston Dynamics' most sophisticated humanoid robot, has already demonstrated remarkable agility and dexterity. However, the new collaboration seeks to further enhance its capabilities by leveraging the RAI Institute's expertise in AI and learning, combined with Boston Dynamics' hardware and software co-design for humanoid robots.

One of the key challenges in robotics is transferring simulation results to real-world robotic hardware. To address this, the partnership will establish a shared reinforcement learning training pipeline for Atlas, enabling the robot to learn and adapt to new environments more efficiently. By training policies in simulation and refining them on physical hardware, the teams aim to generate novel, robust, and practical locomotion behavior for Atlas.

The collaboration also focuses on improving whole-body loco-manipulation, allowing Atlas to manipulate objects and fixtures in conjunction with locomotion. This will significantly enhance the robot's utility in various environments, making it more adaptable to different tasks and applications.

Additionally, the partnership will explore full-body contact strategies, such as dynamic running and full-body manipulation of heavy objects. This will necessitate close coordination between the arms and legs, enabling Atlas to generate such behavior during complex contact events without imposing strict requirements.

By working together, Boston Dynamics and the RAI Institute hope to streamline the process of achieving new skills for Atlas, expanding its skillset and making it a more valuable tool in people's lives. The organizations will periodically release updates on this collaborative work, including publications and demonstrations with Atlas.

In conclusion, the partnership between Boston Dynamics and the Robotics & AI Institute, led by former CEO Marc Raibert, is poised to accelerate the development of reinforcement learning for the Atlas humanoid robot. By combining their expertise and resources, the organizations aim to bridge the sim-to-real gap, improve whole-body locomotion and manipulation, and create a more capable, adaptable, and useful robot for various applications.