waymo-research/waymax

Health v2 · maintenance only

Flags: no_releases no_license

Adoption not part of the score

1096 stars · 132 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Waymax is a lightweight, multi-agent autonomous driving simulator built entirely in JAX, based on the Waymo Open Motion Dataset. It supports closed-loop planning simulation, sim agent research, and open-loop behavior prediction using bounding-box object representations, and runs efficiently on GPUs and TPUs.

Use cases

• simulate autonomous driving scenarios for research
• train and evaluate planning policies in closed-loop simulation
• run behavior prediction experiments on the Waymo Open Motion Dataset
• build learned sim agents for driving simulation
• benchmark driving policies with common metrics and RL interfaces
• accelerate driving simulation on GPUs and TPUs with JAX

When to choose

• you need a fast, JAX-accelerated multi-agent driving simulator
• your research uses the Waymo Open Motion Dataset
• you want closed-loop evaluation for planning or sim agents
• you need GPU/TPU-distributed simulation with RL environment adapters

When to avoid

• you need raw sensor data simulation rather than bounding-box abstractions
• you want a commercial-use simulator (license is non-commercial only)
• you need a full autonomous driving stack with perception and sensor simulation
• you don't work in Python or the JAX ecosystem

Facets

library · maturity active

simulation machine-learning reinforcement-learning data-science autonomous-vehicles machine-learning robotics simulation python jax autonomous-driving waymo-open-motion-dataset behavior-prediction closed-loop-simulation sim-agents motion-forecasting gpu linux macos

Member repositories

For agents

markdown · JSON · MCP: product_card(name="waymo-research/waymax")