The joint effort, initiated in Massachusetts, also involves developing a reinforcement learning research kit for Boston Dynamics' Spot robot. The current focus is on transferring simulation-based learning to real-world applications and enhancing Atlas's movement and interaction with physical environments. This includes dynamic running and full-body manipulation of heavy objects, which require synchronized leg and arm movements. The bipedal form of Atlas presents unique challenges and opportunities compared to Spot, with activities influenced by various forces like balance and motion.
Key takeaways:
- Boston Dynamics has partnered with the Robotics & AI Institute to enhance reinforcement learning for its Atlas humanoid robot.
- The partnership aims to improve Atlas's ability to learn new tasks through reinforcement learning, which is traditionally time-intensive but now aided by effective simulation.
- The collaboration focuses on transferring simulation-based learning to real-world settings and improving Atlas's movement and interaction with physical environments.
- The partnership is part of broader efforts by Boston Dynamics to explore advanced technologies, supported by Hyundai, which owns Boston Dynamics and funds the Institute.