Smart Streets

How a System Used AI to Improve Pavement Management

The Client

The client was a startup with the idea to create an AI-powered infrastructure intelligence platform. The product aimed to help use modern technology to transform how they assessed, planned, and managed road infrastructure.

As roads continue to age, public agencies must keep pavement in good repair while staying within their limited budgets. Deciding what roads to prioritize for repairs has historically used manual inspections and isolated documentation. Often, repairs rose to the top of the to-do list only after massive repairs were needed rather than catching those issues at their beginning stages.

The client aimed to create an application using the power of AI to proactively repair problems before they devolved into massive issues requiring more extensive and expensive repairs. The client’s data-driven platform provided support for the full lifecycle of infrastructure decision-making from assessment to execution.

How It Works

Gather pavement data

Analyze pavement decline

Optimize maintenance plans

Complete and monitor work

The Idea

The client company has worked in digital road accounting and damage rating for a long time. Their work covers specific regions and the specific roads they are responsible for maintaining.

Before our project, they used various office products to organize the work of drivers, managers, and road accounting and damage scoring experts. They used a specific table format for this, along with other non-centralized tools. Most of the company’s staff were drivers who captured roads with a phone, road experts who rated damage, and group managers.

The idea behind this project was to build a working tool that would help rate road damage using AI and would streamline the process of taking and uploading photos.

Goals

The main goal was to build an AI-based system for road rating and accounting.

The system consists of two main parts: a web app with all the tools for managers, and a mobile app that drivers use to capture roads in the field.

Secondary goals were to streamline working processes and to give both company managers and drivers more information about the current state of the work.

AI-based scoring also means they will potentially need fewer people to review road photos, as the process is largely automated.

Technology

Details

Type of Project

Both mobile and web

User Stories

System for road capturing and AI-assisted rating

Purpose

Optimize the road rating process

Team

When the project began, the team consisted of a project manager, two Ruby On Rails developers on the web, one Flutter developer on mobile, and two manual QA engineers.

Throughout the initial phases, Swan Software Solutions provided a senior Ruby On Rails developer and a senior mobile developer to help with consultations, initial technical decisions, stack choices, and code review.

As a Ruby on Rails developer, the client company’s Chief Technology Officer collaborated with Swan’s team who worked on a full-time-equivalent basis.

Timeline

As as a full-time-equivalent project, Swan’s team became a part of the client company’s technology team. This allowed the client company and Swan’s team to work together seamlessly from the beginning of the project to the end.

Swan’s incorporation and collaboration allowed the client company to speed up the development process that would have been slower if only completed by in-house developers.

Project Start: 11/18/2024 – Project Complete 06/30/2026

The Problem

Before our project, they had a different, more expensive workflow: company staff recorded videos with a phone camera, and those videos were sent to a contractor who cut freeze-frames from them and charged per video file.

The frames were then analyzed by road specialists and added to scoring tables, mostly manually. This was slow and expensive, and because of the photo quality, it was hard to score roads correctly.

Drivers also had to use a regular phone map and add custom markers to keep track of which roads had already been captured and which had not.

The Focus

On the web side, we had to build a custom parser for road data tables that contain not only rating data but also map data in a specific industry format.

The developers also had to create a map data optimizer to make it work smoothly with a large amount of data on display.

The web team also had to add AI photo analysis using a custom prompt that helps the AI score roads correctly within the road rating system.

On mobile, we have one particularly complex road-capturing screen that has to take photos, navigate the user, handle map markers, and support a lot of other minor features.

This screen, along with the extensive work on the camera and its settings (the app handles double photo capture, focus, exposure, distance between photos, etc.), was what was most specific to this project.

Solution

Swan delivered web and mobile solutions that addressed the company’s complex data rendering and road-capture needs. There were a lot of small things done that, taken together, helped solve the company’s core problem by streamlining field data collection.

We developed a solution that provides:

Manager’s side

• seeing which roads are already captured and which are not;

• dividing roads into assessments for a group of drivers;

• seeing AI ratings of captured roads and changing them if needed;

• seeing statistics on road status, scores, etc.

Driver’s side

• capturing photos, navigating on the map, and seeing in real time what the driver has already covered;

• uploading photos directly to the web for AI analysis, or saving them to external storage for a bulk upload later when the internet connection is poor.

We also added a parser for their custom table format that uploads all previously completed progress into the system so they can keep working from it.

Implementation

Conclusion

Swan Software Solutions left this project with the client company having a robust web and mobile app with a lot of tools for their staff to track, add, and score roads.

Our system resolves many of their issues with photo quality and capture cost, and the AI-based rating system takes a lot of the workload off the road experts.

The client continues to use and enhance the application with a smaller internal team now that Swan’s work is done.

Schedule an Appointment to Start your Project

More Case Studies