Nowadays, you find ‘all robotics in one place’ – the original slogan of this website – on twitter. Our twitter account, @planetrobotics, will only occasionally come up with own tweets. Most of the time you will find us retweeting the most interesting robotics-related stuff we can find for our >13K followers. You should definitely follow us!

If you want to know more right now you have to read 😉
Here is an interview @arnenordmann did with me about @NotMyRobots for @robotik / @planetrobotics

Tired of Atari? Try to run RL on real-robots and participate in the, an official NeurIPS21 challenge. Submit your jobs to the robot like to a cluster and extend any of the baselines for winning prizes! Starting now!

Mars Helicopter Lands Safely After Serious In-Flight Anomaly

Happy to share a project we've been working on for past 2+ years: MT-Opt Multi-task RL at scale on real robots works! We amortize cost of robot learning over multiple tasks including instance and indiscriminate grasping, placing, aligning and rearranging🧵

Google AI@GoogleAI

Presenting two new approaches to robotic #ReinforcementLearning at scale — MT-Opt, an RL system for automated multi-task data collection and training, and Actionable Models, a system for multi-task goal-conditioned RL. Learn more below

We collected >800k episodes of multiple tasks which was a big challenge in itself: we trained CNN success detectors to determine rewards at scale, built automated reset mechanism to collect data without supervision and we use solutions of easier tasks to bootstrap harder tasks.

With #WaymoVia, we’re building on our past decade of learnings developing autonomous driving technology to deploy a reliable, fully autonomous trucking solution. More on how this experience sets us apart in @TransportTopics:

The Cheetah's Fluffy Tail Points The Way for Robots With High-Speed Agility

Load More...