Getting your Trinity Audio player ready...
|
The rapid development of AI is not confined to specific geographic regions; it is a global phenomenon. Advances in artificial intelligence are taking place on the ground and soaring to new heights in the aviation sector. AI has been making significant strides within the aviation industry, transforming how aircraft are operated, monitored, and maintained.
Recently, researchers from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) have unveiled the Air-Guardian, an innovative news system that acts as a proactive copilot, forming a collaborative partnership between humans and machines centred around understanding attention.
The system employs eye-tracking technology for humans, while the neural component relies on “saliency maps” to pinpoint the direction of attention. These maps serve as visual guides, emphasising crucial areas within an image and aiding in comprehending intricate algorithms’ behaviour. Unlike traditional autopilot systems that only intervene during safety breaches, Air-Guardian utilises these attention markers to identify early signs of potential risks.
The significance of this system extends beyond aviation, potentially finding applications in automobiles, drones, and a broader range of robotics. MIT CSAIL postdoc Lianhao Yin explained that the system’s cooperative layer and the entire process can be trained using a dynamic neural network model that maps attention and adaptability. This adaptability ensures a balanced partnership between humans and machines.
Field tests demonstrated that both the pilot and the system made decisions based on the same raw images while navigating to the target waypoint. Air-Guardian’s success was measured based on cumulative rewards earned during flight and the efficiency of reaching the target waypoint. The system effectively reduced flight risk levels and improved the success rate of navigation.
Ramin Hasani, an MIT CSAIL research affiliate and the inventor of liquid neural networks, emphasised the innovative human-centric approach of the air guardian, highlighting its dynamic and adaptive nature, which complements human judgment, enhancing safety and collaboration in aviation.
The core strength of Air-Guardian lies in its foundational technology. It utilises an optimisation-based cooperative layer that incorporates visual attention from humans and machines and liquid closed-form continuous-time neural networks (CfC), known for their proficiency in understanding cause-and-effect relationships. Additionally, the VisualBackProp algorithm identifies the system’s focal points within an image, ensuring clarity in its attention maps.
To promote future widespread adoption, refinement of the human-machine interface is essential. User feedback suggests that an indicator, such as a bar, is a more intuitive way to signify when the guardian system takes control.
Air-Guardian represents an advancement in aviation safety as it is providing a dependable safety net for situations where human attention may falter. It exemplifies the relationship between human expertise and machine learning to augment human capabilities in challenging scenarios and reduce operational errors, as noted by Daniela Rus, Andrew (1956) and Erna Viterbi, Professor of Electrical Engineering and Computer Science at MIT and director of CSAIL.
Stephanie Gil, an assistant professor of computer science, also underscored the potential for earlier interventions and improved interpretability through visual attention metrics, showcasing how AI can work harmoniously with humans and foster trust in collaboration. The integration of AI into aviation is a testament to the versatility and adaptability of AI technologies across various domains.