ROWMow Sim: Purdue and INDOT collaborate for safer highway mowing

Roadside mowing is inherently dangerous, time-consuming and costly.

The Indiana Department of Transportation (INDOT) spends approximately $18 million annually maintaining more than 40,000 acres of roadside vegetation. Contractors must often operate just feet from high-speed traffic, sometimes crossing into travel lanes and navigating around obstacles hidden in tall grass. 

To address the challenges facing mower operators, INDOT partnered with Purdue on a research project funded through the Joint Transportation Research Program. Led by John Evans, assistant professor of agricultural and biological engineering (ABE) in Purdue’s College of Agriculture and mechanical engineering (by courtesy) in the College of Engineering, the project explores whether autonomous mowing systems could improve safety and efficiency.

Autonomous mowers were a promising option, but INDOT needed a way to determine whether such systems could safely navigate complex roadside conditions before deploying them on actual highways.

“Autonomous mowers for residential and light commercial use have been available for a while, but highway mowing presents fundamentally different challenges with highly variable terrain, tall vegetation that obscures obstacles and the need to operate safely next to high-speed traffic,” Evans said. “A high-fidelity virtual environment allowed us to test the safety of autonomous systems without putting people and equipment in harm’s way.”

To build an effective testing environment, the research team first needed comprehensive data about actual mowing operations.

INDOT provided detailed information, including GPS-located vector data with road and sign locations and high-resolution light detection and ranging (lidar) scans to improve fidelity for localized areas. Purdue researchers mounted time-lapse cameras on INDOT contractors’ mowing equipment to gather imagery and performance data. Over two years, they were able to produce hundreds of thousands of images and terabytes of data.

Graduate students in the Evans Lab mapped obstacle density patterns along roadsides. Using machine learning models, they analyzed datasets to identify common challenges, such as obstacles or instances when equipment entered travel lanes.

The analysis revealed that mowers encroached on highways more frequently than expected, and vegetation height often concealed hazards.

“My lab has done a lot of work with cosimulation and codevelopment frameworks, using virtual environments to inform the physical world and vice versa,” Evans said. “This allows us to speed up development on projects where real-world testing is limited.”

The collaboration produced ROWMow Sim, a virtual digital twin of real Indiana roadside locations in a scalable testing environment. The program is built on Unreal Engine 5.1, which is the industry standard for physics in games like Fortnite and Gears of War.

“We were able to download GPS-referenced features, road signs, guardrails and the road itself and pair it with lidar scans, then pull it into Unreal Engine,” Evans said. “This gave us a virtual environment where we could safety test mowing conditions.”

In ROWMow Sim, virtual mowers are equipped with sensors that simulate real-world challenges, such as GPS signal loss and environmental noise. This data-driven approach ensures that virtual testing accurately reflects the complexity and unpredictability of actual roadside mowing conditions.

The virtual world met the real world when researchers validated ROWMow Sim’s accuracy at a controlled test site. INDOT provided guardrails, signage and obstacles to create identical physical and virtual situations.

“We were able to test both the simulated and real environment with the same conditions to see how they aligned,” Evans said.

With fidelity confirmed, the safety reports from the mirrored simulated and physical tests provided metrics that developers and INDOT can use to evaluate autonomous mowing system reliability.

Graduate students in the Evans Lab led simulation development, algorithm validation and safety reporting throughout the project. Evans emphasized the value of collaborating with industry partners.

“Giving students a chance to engage with partners and build relationships that could lead to career opportunities after they’ve finished their degree means a lot to me,” he said.

The project also benefitted from cross-departmental collaboration with Greg Shaver, director of Ray W. Herrick Laboratories and the Reilly Professor of Mechanical Engineering. The team worked closely with Matt Kraushar, roadside maintenance specialist at INDOT, to provide updates and gather feedback on the project.

While the autonomous mower pilot project phase has concluded, Purdue and INDOT continue to explore ways to improve highway maintenance safety. ROWMow Sim is publicly available to download on GitHub and has already been used for other research into autonomous systems.

“We’ve used the ROWMow environment for other projects,” Evans said. “It’s a useful tool that we hope other people will be able to use to assess and mitigate risk in autonomous decision-making.”