Agents
Bluebird's AI agents can manage complex air traffic scenarios using a range of complementary approaches. All are developed and rigorously tested within Bluebird's digital twin.
Research Approaches
Rules-Based
Structured controllers that encode expert knowledge directly — fast, interpretable, and a meaningful baseline for evaluating more complex approaches.
Optimisation
Methods that search for optimal actions within a defined problem space. Falcon, an optimisation-based agent, controls simulated sectors directly and forms a key part of the Bluebird agent toolkit.
Reinforcement Learning
Agents that learn through interaction with the digital twin, building strategies through simulated experience. Well-suited to scenarios too complex to specify by rule alone.
Search-Based Methods
Agents that use forward search to find good action sequences within the simulator. Complementary to learning-based approaches, and well-suited to structured planning problems.
And more — the platform is agent-method agnostic
Bluebird's digital twin uses a standardised gym-style interface, meaning any agent methodology can be plugged in and evaluated on the same scenarios. Explore the open-source platform to try your own approach.
Read Our Agent Research
Explore publications on agent research across multiple methodologies in air traffic control.
- A Future Capabilities Agent for Tactical Air Traffic Control
AIAA SciTech Forum, 2026
- Human-in-the-Loop Testing of AI Agents for Air Traffic Control with a Regulated Assessment Framework
AIAA SciTech Forum, 2026
- Online Action-Stacking Improves Reinforcement Learning Performance for Air Traffic Control
AIAA SciTech Forum, 2026
- A Probabilistic Digital Twin of UK En Route Airspace for Training and Evaluating AI Agents for Air Traffic Control
AIAA SciTech Forum, 2026