Uploaded March 2025 | Updated September 2026, 3 weeks ago

---

We are happy to release an update for the Aerial Gym Simulator.

Code: github.com/ntnu-arl/aerial_gym_simulator

The new update includes support for multi-linked embodiments, with fixed, reconfigurable (active), and soft (passive) joints for simulated multirotors. A faithful simulation model of a compliant robot Morphy(youtube.com/watch?v=C6l7Vklbc9k) and models for active reconfigurable multirotor platforms are added!

We provide scripts for training:
- State-based RL policies for position setpoint tracking including end-to-end methods
- Vision (depth) based RL-policies for navigation of cluttered environments
- Joint-aware policies for soft drones that minimize oscillations of compliant joints
- Joint and motor control policies for reconfigurable systems for robot-shape and position control

Newer capabilities to query face and vertex indices and surface normal information are added to the rendering framework alongside the capability to query user-defined vertex-level annotations from the environment.
Scripts to deploy trained policies on real robots are provided alongside network model files for easy reproducibility.