RTD Models Now Integrated with Python Learning Libraries

We’re thrilled to announce a major milestone—RTD’s simulation model libraries are now successfully integrated with Python-based machine learning frameworks. This opens the door to powerful new possibilities in AI training for aerospace applications.

To showcase the potential, our team has developed several example agents trained using reinforcement learning and neuroevolution techniques:

🛩️ Dogfight AI Agent

Trained in a custom Python Stablebaselines compatible Gym environment, this agent engages in a 1-on-1 aerial combat. Featuring IR missiles and flare countermeasures, the scenario leverages FixedWingLib CGF for flight dynamics, CML Add-on for maneuvering, and EWAWS for realistic weapons systems. The reward model is based on combat geometry and other tactical state variables. Curriculum training was used to accelerate the learning.

🛰️ Drone Interceptor AI Agent

In another custom Gym environment, we trained a quadcopter drone to defend a ground target by intercepting an attacker midair. Built on RotorLib CGF, this simulates highly dynamic physics-based flight for both the attacker and defender drones, allowing the AI to learn precise interception strategies.

This integration paves the way for exciting future developments in autonomous systems, AI training workflows, and simulation-based research.

Interested? Please contact us here.

---

Interested in AI topics? Join the RTDynamics AI Newsletter here