1. Boston Dynamics produced two major robotics news cycles last week, including the electric Atlas announcement and a farewell to the original hydraulic Atlas.
2. The company promotes transparency in showcasing the ups and downs of robots, emphasizing the importance of learning from failures.
3. Falling is a crucial aspect of robot development, with robots like Spot and Digit learning to fall well and get back up again to ensure uninterrupted automation.
Boston Dynamics made waves last week with the announcement of the new electric Atlas robot, which garnered millions of views in a short amount of time. The company also bid farewell to the original hydraulic Atlas, reflecting on its decade-long journey from a DARPA research project to a highly capable bipedal robot. The company’s transparency in showcasing the robot’s falls and successes is a welcome reminder of the effort that goes into creating these robots and the importance of learning from failures.
Boston Dynamics’ CTO emphasized the importance of robots falling during real-world tasks as a way to learn and improve their capabilities. The ability to withstand falls and get back up without breaking is essential for robots like Spot, which are used in various applications and environments. The company has learned valuable lessons from Spot’s real-world experiences, leading to improvements in resilience and safety.
Agility Robotics also addressed the issue of falling in robots, highlighting the role of arms in protecting the robot and assisting in recovering from falls. The company uses reinforcement learning to help fallen robots get back up, emphasizing the importance of learning from failures. In industries where humanoid robots are integrated into existing workflows, the ability to recover from falls autonomously is crucial for uninterrupted automation.
The concept of falling as a learning opportunity for robots is key to the advancement of bipedal robotics. Boston Dynamics and Agility Robotics have embraced the idea that falling is a part of the learning process and have implemented strategies to help their robots recover and improve from these experiences. This approach is essential for the future of robotics in real-world applications.