The robotics team at Google have developed a ping pong robot based on the i-S2R platform that is able to sustain a 340 hit rally. Iterative-Sim2Real essentially allows a robot to hold rallies of over 300 hits with a human player, while GoalsEye enables learning goal-conditioned policies that match the precision of amateur human table tennis players.
What is Iterative-Sim2Real? Basically a training platform that teaches a robot to carry out a rally with a human for as long as possible. This simulation-based model alternates between real and simulated modes until it was good enough to sustain rallies. In each iteration, the human behavior model and the policy are refined, but the ultimate goal is to build a robot trainer that could quickly adapt its play style to the skill level of its human opponent. However, it may be a while until we can teach human brain cells to play table tennis.
- Unlocked Android 5G phone gives you the flexibility to change carriers and choose your own data plan[1]; Pixel 6 is fast, smart, and secure, and...
- The powerful Google Tensor processor is the first processor designed by Google and made for Pixel; it keeps your phone fast, your games rich, and your...
- Pixel’s 50 megapixel rear camera captures 150% more light for photos with richer colors and more detail[2]
The synthesis of techniques is crucial. The policy’s objective is to return a variety of incoming balls to any location on the opponent’s side of the table. A policy trained on the initial 2,480 demonstrations only accurately reaches within 30 cm of the goal 9% of the time. However, after a policy has self-practiced for ~13,500 attempts, goal-reaching accuracy rises to 43%,” said Avi Singh, Research Scientist, and Laura Graesser, Research Engineer, Robotics at Google.
