SimTooReal

Train and deploy robot policies with sim-to-real calibration and automated orchestration.

Freemium - Free plan — no credit card required. Paid tiers likely available for increased compute, fleet deployment and advanced features (see site for details).

Visit Website

About SimTooReal

SimTooReal is a robot policy training and sim-to-real deployment platform for Isaac Lab and MuJoCo. It monitors 28+ real-time training signals, auto-detects and auto-tunes failed or stalled runs, and closes the sim-to-real gap using Bayesian domain-randomization calibration derived from hardware telemetry. Launch and manage training via a web dashboard, CLI, or CI hooks, then deploy safely with physics-aware safety gates, canary rollouts and one-click rollback. Free plan available — no credit card required.

Key Features

SSH-dispatched training orchestration for Isaac Lab and MuJoCo with 4,096+ parallel environments
Monitoring of 28+ real-time training signals (reward, entropy, KL divergence, FPS, GPU utilization)
Automated failure detection and auto-tuning: detects entropy collapse, KL spikes, and reward plateaus
Bayesian domain randomization calibration using hardware telemetry to close sim-to-real gap
Physics-aware deployment safety gates, operator preflight checks, canary rollouts and rollback

Who is this for?

Robotics researchers and engineers, reinforcement-learning practitioners, and teams deploying robot policies to hardware.

Comments (0)

No comments yet

Be the first to share your thoughts!

Stats

Upvotes0

Comments0

Pricing

Freemium

Free plan — no credit card required. Paid tiers likely available for increased compute, fleet deployment and advanced features (see site for details).