deBilla/robot-fleet

Domain randomisation for sim-to-real robustness 11 days ago

enhancement help wanted area: ml

1

Robot fleet platform with a cloud brain: simulated MuJoCo humanoids stream telemetry over gRPC/Kafka, the fleet's own experience trains walking policies offline (CQL/PPO on Kubeflow + Argo), and models roll out through a gated canary. Go, Python, Temporal, k3s.

Go

#argo-workflows#fleet-management#golang#grpc#helm#humanoid#k3s#kafka#kubeflow#kubernetes#mlops#mujoco#offline-rl#python#reinforcement-learning#robotics#simulation#stable-baselines3#temporal

Walking policy plateaus at ~80 ticks: reward favours lunging over balancing 12 days ago

enhancement help wanted area: ml

deBilla/robot-fleet

1

Robot fleet platform with a cloud brain: simulated MuJoCo humanoids stream telemetry over gRPC/Kafka, the fleet's own experience trains walking policies offline (CQL/PPO on Kubeflow + Argo), and models roll out through a gated canary. Go, Python, Temporal, k3s.

Go

#argo-workflows#fleet-management#golang#grpc#helm#humanoid#k3s#kafka#kubeflow#kubernetes#mlops#mujoco#offline-rl#python#reinforcement-learning#robotics#simulation#stable-baselines3#temporal

Remove or archive the legacy Go kinematic simulator 12 days ago

good first issue area: simulator

deBilla/robot-fleet

1

Robot fleet platform with a cloud brain: simulated MuJoCo humanoids stream telemetry over gRPC/Kafka, the fleet's own experience trains walking policies offline (CQL/PPO on Kubeflow + Argo), and models roll out through a gated canary. Go, Python, Temporal, k3s.

Go

#argo-workflows#fleet-management#golang#grpc#helm#humanoid#k3s#kafka#kubeflow#kubernetes#mlops#mujoco#offline-rl#python#reinforcement-learning#robotics#simulation#stable-baselines3#temporal

k3s-setup.sh disk check only works on macOS 12 days ago

bug good first issue area: infra

deBilla/robot-fleet

1

Robot fleet platform with a cloud brain: simulated MuJoCo humanoids stream telemetry over gRPC/Kafka, the fleet's own experience trains walking policies offline (CQL/PPO on Kubeflow + Argo), and models roll out through a gated canary. Go, Python, Temporal, k3s.

Go

#argo-workflows#fleet-management#golang#grpc#helm#humanoid#k3s#kafka#kubeflow#kubernetes#mlops#mujoco#offline-rl#python#reinforcement-learning#robotics#simulation#stable-baselines3#temporal

evaluate_policy: failure_reason ignores the episode-length condition 12 days ago

bug good first issue area: ml

deBilla/robot-fleet

1

Robot fleet platform with a cloud brain: simulated MuJoCo humanoids stream telemetry over gRPC/Kafka, the fleet's own experience trains walking policies offline (CQL/PPO on Kubeflow + Argo), and models roll out through a gated canary. Go, Python, Temporal, k3s.

Go

#argo-workflows#fleet-management#golang#grpc#helm#humanoid#k3s#kafka#kubeflow#kubernetes#mlops#mujoco#offline-rl#python#reinforcement-learning#robotics#simulation#stable-baselines3#temporal

Log reward components, torso height and speed in test-track evaluation 12 days ago

enhancement good first issue area: ml

deBilla/robot-fleet

1

Robot fleet platform with a cloud brain: simulated MuJoCo humanoids stream telemetry over gRPC/Kafka, the fleet's own experience trains walking policies offline (CQL/PPO on Kubeflow + Argo), and models roll out through a gated canary. Go, Python, Temporal, k3s.

Go

#argo-workflows#fleet-management#golang#grpc#helm#humanoid#k3s#kafka#kubeflow#kubernetes#mlops#mujoco#offline-rl#python#reinforcement-learning#robotics#simulation#stable-baselines3#temporal

Open Issues Need Help