126 Savilionis’s team tunes the AI model to consider different parameters in its priority ranking. They asked it to rank streets based on their importance for pedestrian traffic. Before that, they tuned it to account for routes of cyclists and commuters. The AI model notes, for example, how close a street is to schools and hospitals, and other places where many people need to travel. “A street with really bad conditions might rank low on the metric because of its distance from healthcare or educational institutions,” Savilionis said. He noted one street that had a high score for urgency due to its proximity to a playground. This information helps expedite repairs and can be shared with the public, increasing transparency and accountability. Enhancing a Shared Awareness of the City In addition to real-time analysis, Vilnius’s digital twin helps officials understand how city operations might evolve over time. The City of Vilnius, Lithuania’s capital, employs drones and AI functions to run the city in the present and decide how to change it in the future. Creating a Real-Time Digital Twin Vilnius complements its twin with drones to enable real-time analysis. The city’s fleet of four drones gathers imagery and other data that can be projected onto maps and analyzed directly or via AI models. Savilionis pulled up another map, this one with municipal trash receptacles highlighted. As with snowy streets, the AI had been trained to recognize full trash bins. Those in current need of emptying were highlighted in red. Maps inside the digital twin serve more than one purpose. In addition to helping the city manage municipal functions, they provide a record that lets officials compare services over time. The public-facing versions also provide transparency for Vilnius residents. Training an AI Model to Prioritize Attention The relative ease with which drone footage becomes operational intelligence belies the massive effort used to train the AI model. Thousands of photos are used to teach the algorithm what to look for, a process known as deep learning. The deep learning models then ingest the latest drone imagery to return answers. Once trained, the AI can spot priorities for service. These areas appear on dashboards for city leadership and guide the work of relevant city departments. For example, the city can assess road conditions by using the AI to highlight cracks, potholes, and other damage that may require attention. The model can also recognize where traffic is backed up, where cars are parked illegally, and where large snow loads have accumulated on roofs. “That’s especially useful for buildings that might be hard to access by foot,” Savilionis said.
RkJQdWJsaXNoZXIy MjA2NTE0Mw==