Applies to v0.1.0.

Gym interface

Planned. A Gym-compatible interface is on the roadmap but not available yet. This page will document the environment API once it lands.

The goal is a Gym-style step / reset environment, so standard reinforcement-learning stacks can train against Nano-sim without custom glue.

Intended shape

• Observation — camera frames, LiDAR point clouds, and GNSS/IMU state.
• Action — continuous steering and throttle commands.
• Reward — task-specific, e.g. lane-following progress and stability.