Scaling up intelligent transportation pilots is a challenge – but without working out post-pilot funding, ownership and public engagement, you will be left with just “a really impressive science project”.
That was one of the key conclusions from a webinar moderated by ITS International editor Adam Hill with experts from OnLogic and Intel, which looked at how to maximise your existing assets, ensure the right hardware is in place and meet Build America Buy America (BABA) requirements.
The speakers were Sheldon Sun, VP product engineering, OnLogic, and Tony Abuta, technologist – smart cities and intelligent transportation, at Intel, and they started by looking at the hurdles which agencies face when trying to scale up their pilot projects.
Click here to see the whole webinar.
For Sun, there are three basic traps to fall into, the first of which is “running a perfect pilot”.
“Obviously when you run a pilot, you’re in an environment where everything’s perfect: your cameras are tuned perfectly, the intersections are aligned perfectly, there’s no construction… you get the perfect data, you can hand-tune all of the algorithms in that specific scenario.”
The second trap is to think that all existing infrastructure needs replaci
ng – but that means you’re changing a lot of variables at the same time.
“The goal here is to reuse as much, and to change as little variable as you can to just get that repeatability and ruggedness from what’s already existing.”
The third one is what Sun calls “the hardware trap”: “This is where you look at a system in the field that may be a little bit non-optimal – that may mean a lot of dust, a lot of heat – and a system normally needs to breathe. And because of that, if you run into a system that has a fan, or doesn’t have adequate cooling, you run into a really messy environment where you have hardware failures.”
“So in order to actually survive this mass deployment initiative, you’ll need a rugged system; you’ll need a lot of headway in terms of environmentals – that would be thermals and also allowing your system to actually breathe while it’s working.”
Tony Abuta at Intel took up the theme, highlighting pilots which look “amazing when they’re put in a bubble”.
“We had near-miss detections that worked really awesome; we had digital twins, we had real-time analytics; but then they didn’t scale. That’s always been a frustration – you’re just spinning your wheels. And it’s not that technology failed – it’s because no-one had answered the real question, which is ‘who is funding these 300 devices that are going to scale out across the city?’ Pilots, they often run on innovation grants. But those scaling capabilities require capital budget: you know, procurement needs to be aligned, you have to have the ITS security reviews, and then you have to have that clear operational ownership.”
Public perception is also important in determining the success of a project, “especially when you’re deploying AI and all these cameras out in the field”.
“There’s a real pushback around cameras and AI – sometimes it’s misinformation, sometimes it’s privacy concerns, and sometimes it’s just political, right? If leadership is not prepared to explain what the system actually does… it’s not a technical problem that’s a problem scaling this.”
“We have to make sure that institution is ready for that deployment, we have that funding, we have that ownership, and also we have that public communication so that the public knows when we’re installing these AI systems in the cabinets by the traffic intersections, that these things are there to help them to save lives and also to improve traffic flow.”
“We really need to have that stakeholder alignment,” he concluded. “So if a pilot is not really engineered for that reality from day one, it just becomes a really impressive science project.”