Bitbots
A compact, lightweight legged robot product that uses a reinforcement-learning-driven locomotion control scheme, positioned along a route close to consumer-grade quadruped robots.
Timeline (last 90 days)
The Bitbots project page is https://bitbots.endresgames.com, and free invitations can be obtained via comments or direct messages.
Bitbots is currently in a closed alpha.
In Bitbots, players decide the robot's body and joint configuration, and the policy network learns how to control it.
During matches, Bitbots robots act autonomously and players cannot intervene mid-match.
The developers stress that the "brain view" displays recorded values from the actual policy, not decorative animation.
Developer Bitbots is a browser game in which players assemble robots from parts and then train them to walk through reinforcement learning.
The Bitbots game provides a "brain view" that shows the policy network's inputs, hidden-layer activations and motor outputs in real time.
The Bitbots teaser shows that it is taking a miniaturized, lightweight route, close to the product logic of Xiaomi's CyberDog and Unitree's Go series, namely using consumer-grade hardware plus reinforcement learning to bring legged robots into a mass-producible price band.
Learning to walk in 93 hours only shows that basic locomotion capability has taken shape; there is still a long way to go before stable stair climbing, disturbance resistance, and long battery life, and training duration does not equal product maturity.
As for Bitbots, what the developer community will really ask is whether its training framework is open source and whether its sim-to-real solution is reproducible.
1 more timeline events
All Facts (11)
The Bitbots project page is https://bitbots.endresgames.com, and free invitations can be obtained via comments or direct messages.
50%UnverifiedBitbots is currently in a closed alpha.
50%UnverifiedIn Bitbots, players decide the robot's body and joint configuration, and the policy network learns how to control it.
50%UnverifiedDuring matches, Bitbots robots act autonomously and players cannot intervene mid-match.
50%UnverifiedThe developers stress that the "brain view" displays recorded values from the actual policy, not decorative animation.
50%UnverifiedDeveloper Bitbots is a browser game in which players assemble robots from parts and then train them to walk through reinforcement learning.
50%UnverifiedThe Bitbots game provides a "brain view" that shows the policy network's inputs, hidden-layer activations and motor outputs in real time.
50%UnverifiedThe Bitbots teaser shows that it is taking a miniaturized, lightweight route, close to the product logic of Xiaomi's CyberDog and Unitree's Go series, namely using consumer-grade hardware plus reinforcement learning to bring legged robots into a mass-producible price band.
50%UnverifiedLearning to walk in 93 hours only shows that basic locomotion capability has taken shape; there is still a long way to go before stable stair climbing, disturbance resistance, and long battery life, and training duration does not equal product maturity.
50%UnverifiedAs for Bitbots, what the developer community will really ask is whether its training framework is open source and whether its sim-to-real solution is reproducible.
50%UnverifiedBitbots released its first teaser trailer, and the trailer hints that its robot needs 93 hours of training to learn to walk.
50%