What is Rocky Learns to Walk Live?

It is a public SuperDroid Robots dashboard that runs Rocky’s latest locomotion policy in your browser with MuJoCo WASM and shows live training graphs from recent reinforcement-learning runs.

Is this a live camera feed of a physical robot?

No. The 3D view is a physics simulation. Policy weights and metrics are published from SuperDroid’s training systems; inference and physics run locally in your browser.

What do the graphs show?

They show episode length at done (up to the training max episode length), survival completion rate with EMA, and full-episode rate for the latest training run and the previous two runs. Hover a point for exact values and run start times.

What is the walk-time counter?

It tracks how long Rocky stays upright in your browser session and compares your best to the global best reported by other visitors.

Rocky learns to walk

In-browser MuJoCo policy · live training graphs from SuperDroid Robots

Live dashboard

Loading metrics…

Loading policy…

Run —

This episode 0.0s

Your best 0.0s

World best 0.0s

Starting…

Episode length at done scale 0–1500 (max episode)

Survival completion rate + EMA · 0–100%

Full episode rate push deploy · 0–100%

How this live Rocky demo works

Training happens on SuperDroid’s research stack. As training continues, we publish the latest policy checkpoint and TensorBoard-derived scalars to this dashboard—whether or not the run just improved. Your browser loads the current policy, simulates physics with MuJoCo WASM, and redraws training curves for the live run plus the previous two starts. Leave the page open: graphs and the active policy refresh automatically about every 15 seconds as new publishes land.

Reading the training graphs

Episode length at done — how long Rocky stays upright each episode, capped at the training max (1500 steps).
Survival completion — fraction of the max episode completed, with an EMA trend line.
Full episode rate — how often push-deploy evaluations finish a full episode.

Use Recent zoom to scale the X-axis to the live run (older runs still overlay where they overlap). Switch to Full history for the entire published timeline. Hover any point for the exact value, training step, and when that run started.

About SuperDroid Robots

SuperDroid Robots designs and builds durable mobile robots in the United States. Rocky is a simulated research demo that showcases our locomotion learning work—not a product livestream of a physical humanoid.