The Google research team is using artificial intelligence to teach robots how to move with the agility of real-life animals, like dogs. Accomplishing this feat involved reinforcement learning (RL), where the researchers began by sourcing reference motion clips recorded from an animal and using RL to get the robot to copy those motions. Read more for a video and additional information.
One issue was accounting for the randomness that occurs in real-life, thus they had to introduce this by using physical parameters in the simulation – i.e. changing physical quantities (robot’s mass / friction, etc.). This resulted in a machine learning model that could account small variances and the resulting complications.
- Best Robot Vacuum of 2018 by Digital Trends.
- All-New RoboVac: Re-engineered to be the slimmest* RoboVac (2.85'') but with quiet operation and increased suction power at 1300Pa.
- BoostIQ Technology: Automatically increases suction power within 1.5 seconds when extra vacuuming strength is needed to get the best clean.
We show that by leveraging reference motion data, a single learning-based approach is able to automatically synthesize controllers for a diverse repertoire behaviors for legged robots. By incorporating sample efficient domain adaptation techniques into the training process, our system is able to learn adaptive policies in simulation that can then be quickly adapted for real-world deployment,” said the researchers.