Sponsored by:

• Offline Data-Driven AI Policy Learning: Introducing an offline, data-driven approach that utilises Decision Transformers to learn tactical policies directly from existing expert datasets, bypassing the need for risky online trial-and-error processes.
• Training Time Reduction and Performance Validation: Sharing results that demonstrate how the Decision Transformer-based policy significantly reduces training time from hours to minutes, while maintaining a comparable level of tactical performance to conventional reinforcement learning agents like PPO.
• Rapid AI Model Development and its Applicability to Defence: Providing insight into how the methodology of rapidly generating AI models from data shortens the entire development and deployment cycle, and how this can enhance the reliability and agility of AI decision support tools for human operators in defence systems.