Funding entity Airbus Defence and Space S.A.U
The SIAGEN Project proposes a Deep Learning approach that combines Imitative Learning (IL) and Reinforcement Learning (RL) to overcome the high computational costs and fragility of traditional optimization-based planners, enabling near-optimal, scalable, and robust trajectory generation in dynamic, multi-agent environments. To achieve this, we develop an IL-based pipeline for generating trajectories in complex, cooperative missions (such as SEAD) while respecting kinematic and dynamic constraints, alongside providing technical support to integrate and validate these functionalities within Airbus's FCAS Lab simulation framework. The ultimate goal of this research is to enable faster replanning for a large number of vehicles, ensuring the system can handle complex dynamics and both static and dynamic obstacles where traditional optimization methods fail.
Layman's summary: This project uses Imitation and Reinforcement Learning for scalable, robust multi-agent trajectory generation. We develop an IL pipeline for complex missions and support its integration into Airbus's FCAS Lab, enabling faster replanning where traditional methods fail.
Techniques employed: Reinforcement Learning, Trajectory Optimization, Splines
Airbus_AI_4
