AI and Remote Sensing for Resilient and Sustainable Built Environments: A Review of Current Methods, Open Data and Future Directions
This review critically analyzes 244 studies on AI-driven damage assessment for roads, bridges, and buildings, revealing that data availability and sensing modality constraints—not model superiority—limit operational adoption, and proposes a future research agenda centered on foundation models, multimodal fusion, and physics-informed AI to build more resilient infrastructure.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine the world's infrastructure—roads, bridges, and skyscrapers—as the skeleton and muscles of our modern society. Just like a human body, these structures get tired, develop cracks, and sometimes suffer major injuries from storms, earthquakes, or even just the slow wear and tear of time. For a long time, checking for these injuries meant sending a team of engineers out with clipboards, hard hats, and sometimes even dangerous equipment to climb up and look closely. But what if we could give these structures "super-vision"? This is where Artificial Intelligence (AI) and Remote Sensing come in. Think of AI as a super-smart detective that can look at millions of pictures and instantly spot a tiny crack or a pothole that a human eye might miss. Remote sensing is the detective's telescope and microscope; it's the technology that lets us take pictures of the ground from high above using satellites, or from the air using drones, without ever needing to touch the ground. We care about this because our roads and bridges are getting older, and the weather is getting wilder. If we can't spot the damage early, small cracks can turn into big collapses, costing billions and putting lives at risk.
Now, meet the detectives in this story: a team of researchers who decided to take a giant step back and look at the entire landscape of how AI is currently being used to fix our broken infrastructure. They didn't just build a new tool; they reviewed 244 different studies published between 2018 and 2026 to see what everyone else was doing. Their mission was to answer a simple but tricky question: Are we actually getting better at using AI to save our bridges and roads, or are we just getting better at solving puzzles that don't matter in the real world?
Here is what they found, and it's a bit of a plot twist. You might expect that the "smartest" AI models are the ones winning the day. But the researchers discovered that the AI models aren't actually the stars of the show; the data is. It's like having a world-class chef (the AI) but only giving them potatoes to cook with (the data). No matter how good the chef is, they can only make potato dishes. The study shows that most researchers are stuck cooking with the same few ingredients: a handful of specific datasets that are easy to get, like photos of road cracks taken by drones. Because these datasets are so popular, almost everyone is using the same type of AI (called "CNNs" and "YOLO" detectors) to solve the same tiny problems.
The researchers identified three distinct "neighborhoods" where AI is working, and they all play by different rules:
- The Road Patrol (UAVs and Drones): This is the busiest neighborhood. Since it's easy to fly a drone over a road and take pictures, most studies focus here. The AI acts like a super-fast scanner, zooming in to find potholes and cracks. It's great for spotting small, local problems, but it's like trying to map a whole country by only looking at one street at a time.
- The Bridge Watchers (Satellites and Time): Bridges are tricky. Some researchers use drones to look for cracks, but others use satellites to watch the bridge move over months or years. This is like watching a slow-motion movie of a bridge swaying in the wind. The AI here isn't looking for a crack; it's looking for a pattern of movement that suggests the bridge is about to collapse. This requires a different kind of AI that understands time, not just pictures.
- The Building Guardians (Satellite Aftermath): When a disaster like an earthquake or flood hits, satellites swoop in to take "before and after" pictures of entire cities. The AI here acts like a forensic artist, comparing the two photos to see which buildings are standing and which are destroyed. This is crucial for rescue teams, but it mostly happens after the damage is done.
The big revelation from this paper is that we are hitting a wall. Even though the number of studies has exploded (growing from about 15 a year in 2018 to 61 in 2025), the actual use of these tools by real engineers is still very low. Why? Because the AI is often a "black box." It gives an answer, but it can't explain why. Imagine a doctor telling you, "You need surgery," but refusing to show you the X-ray or explain the diagnosis. Engineers, who are legally responsible for keeping bridges safe, can't trust a tool they can't understand. Furthermore, the AI is trained on data from specific places (like certain countries or specific types of roads). If you take that same AI and try to use it in a different country with different weather or road materials, it often fails. It's like teaching a student to drive only on sunny days in California; when you drop them in a snowy forest in Canada, they don't know what to do.
The paper argues that we don't need "smarter" AI models right now; we need better data and better rules. We need to stop just chasing high scores on computer tests and start testing these tools in the messy, real world. The authors suggest a new way forward: a "three-tier" system where AI helps humans at every level.
- Tier 1: Drones do the quick, close-up checks, and a human engineer makes the final call.
- Tier 2: Satellites scan whole regions after a disaster, and a human analyst interprets the map.
- Tier 3: Satellites watch bridges for years, and a human engineer authorizes repairs based on the AI's long-term trends.
In short, the paper suggests that the technology to save our infrastructure exists, but it's currently trapped in a loop of easy data and unproven models. To truly make our cities resilient, we need to break that loop. We need to feed the AI a wider variety of data, make sure it can explain its reasoning, and build a system where the AI is a helpful assistant, not an autonomous boss. The future isn't about building a robot that can fix a bridge on its own; it's about building a partnership where AI and human engineers work together to keep the world standing tall.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.