Mission: Automated ASRU Adaptation
The current state-of-the-art Automatic Speech Recognition and Understanding (ASRU) solutions in Air Traffic Control (ATC) still suffer from high error rates (over 10%) when encountering new or changing airspaces and airports. This poses a significant risk to technology implementation and leads to operational challenges due to misinterpretation of critical communications. Traditional fixes, such as manually collecting and labeling new data, are prohibitively expensive and time-consuming. Our project, SELF-MADE-ATC, solves this by introducing a fully automated, self-maintaining pipeline that rapidly adapts ASRU systems to new or changing airspaces and airports while continuously evaluating and monitoring its quality. We leverage state-of-the-art Large Language Models (LLMs) and domain-specific contextual information to automatically generate high-quality training data and automate the entire maintenance and adaptation process. This creates a continuous feedback loop that minimizes manual labor and economic risk while ensuring personal data remains fully under the end-users’ control.
Become part of our Journey: We believe in collaborative innovation. If you are interested in learning more about the SELF-MADE-ATC project, contributing operational data or expertise, or simply being part of the journey.
The SELF-MADE-ATC advantage:
Building upon recent technology advancements and industry transfer, SELF-MADE-ATC directly addresses the key challenges in maintaining and adapting ASRU systems for dynamic air traffic control environments. Our solution introduces a cutting-edge, automated maintenance and customization process designed for resilience and accuracy.