I - Increasing the trustworthiness of AI in aviation
Our digital assistants don’t replace humans but collaborate with them, benefiting passengers and industry.
Stefano Riverso, JARVIS project coordinator
As with many industries, integrating artificial intelligence (AI)-enabled digital assistants (DAs) into aviation workflows could improve decision-making and operational performance. Inspired by these potential gains, JARVIS is developing three AI-based DAs: one airborne (AIR-DA), one for air traffic control (ATC-DA) and an airport version (AP-DA). “Our digital assistants don’t replace humans, but collaborate with them, benefiting passengers and industry,” says project coordinator Stefano Riverso, discipline leader autonomous systems at project host Collins Aerospace(opens in new window). Powered by machine intelligence, the AIR-DA is designed to increase flight deck automation, helping pilots manage complex operational scenarios and reducing their workload. The ATC-DA will increase air traffic control automation, benefiting airspace capacity management by offering tactical recommendations, flight plan error correction and short-term traffic forecasting. Finally, by expanding airport automation, the AP-DA will enhance safety, with features such as intrusion detection, and operational efficiency by optimising aircraft turnaround and terminal passenger flow.
Increased safety and performance
The project JARVIS, supported by SESAR JU(opens in new window), is addressing challenges common to all three DAs: ensuring AI decision-making is trustworthy, explainable, safe and ethical; delivering human-centric designs which maximise teamwork with humans; and creating big data and cloud infrastructures for secure and decentralised data management. “We are collaborating with the European Union Aviation Safety Agency(opens in new window), adopting its AI Roadmap(opens in new window) recommendations, and giving feedback intended to push the boundaries of what is achievable,” adds Riverso. Key to this approach is the creation of an Ethics Advisory Board for independent feedback. All three DA prototypes will be validated during 2025, with aviation experts, such as pilots and air traffic controllers, participating in both simulated and real environments. The aim is to provide evidence of enhanced aviation workflows by mid 2026. “Initially, the digital assistants will operate in their own environment, but we envision later integration within multi-agent systems for machine-to-machine collaboration and teaming up with humans when needed,” notes Riverso.
