Boston Dynamics' Spot Gets Handsy

The demonstration showcases Spot manipulating a 15kg (33 lb) tire, which exceeds the arm's maximum lift capacity of 11kg. This is achieved by using the robot's entire body for leverage and control, not just the arm, a technique known as whole-body manipulation. The arm itself has 6 degrees of freedom plus a gripper and can drag up to 25kg. The autonomous behavior is powered by a combination of reinforcement learning (RL) and sampling-based control. At a high level, the system generates dozens of possible future movements simultaneously, using a physics engine called MuJoCo to simulate and select the most fluid and efficient path. This allows Spot to dynamically choose contact points on its arm, legs, and body to move the heavy object. This research originates from the Robotics & AI Institute (RAI Institute), formerly the Boston Dynamics AI Institute. Hyundai Motor Group invested $424 million to establish the center, which is led by Boston Dynamics founder Marc Raibert. The institute focuses on core challenges in robotics, including cognitive AI, athletic AI, and organic hardware design. For the tire demonstration, the complex computations were offloaded to an external computer via WiFi, and an external motion capture system was used to simplify perception. This laboratory setup highlights the significant computational power required for such dynamic control, a key challenge for deploying this level of autonomy in real-world, unstructured environments. This advanced manipulation pushes robotics beyond repetitive manufacturing tasks like welding or palletizing. The ability to handle unstable, heavy objects has direct applications in logistics, warehouse automation, and potentially hazardous material handling, where robots must adapt to unpredictable situations.