From AI-assisted shape recognition and radio message tagging to simply having more eyes and ears on racing happenings, the capabilities of the FIA’s Remote Operations Center (ROC) have grown significantly over the 2023 season. Deputy Race Director Tim Mullion and Head of Single Seater Information Systems Strategy Chris Bentley explain…
How has ROC developed since its inception? What are its main areas of focus?
Tim Marillon: The ROC was launched at the beginning of the 2022 season, three weeks before the Bahrain Grand Prix, and has gone through several stages. The first was getting the technology and architecture right. Therefore, we can say that races 1-4 were observational stages. Races 5 to 8 saw the implementation of higher level technology to enable ROC to actively support Race Control. This included strengthening communication lines between trucks and ROCs, and introducing new equipment and IT tools to provide similar capabilities to ROCs. Tools present in Race Control. Finally, I think the last third of 2023 was actively contributing to problem identification and ongoing regulatory enforcement.
Chris Bentley: This year, the focus was on developing an event management software platform to enable more people and people in remote locations to contribute in a more efficient way. If you look at how we tracked incidents last year, we would look at the video feed and if we saw an issue, we would verbally give the timestamp and car over the intercom and then race control would go and and find it. What we’ve done this year is develop a system that’s more automated so you can pause the video from the ROC, press a button, make sure the data is correct, press the submit button and add the incident to the race control system. It is to do so. .
TM: In our second year, we worked to develop a tool chain for working with people at remote sites and eliminate much of the verbal communication we had in our first year. As strange as it may sound, verbal communication tends to lead to miscommunication and errors.
Track limits continue to be a hot topic in F1. How did ROC contribute to improving the detection process in 2023?
TM: As most people know, track limits really came to prominence at the Austrian Grand Prix, where a number of drivers were penalized after the race and around 1200 potential track limit violations were investigated. I did. The approach has changed significantly between then and now. As an example, at the recent Qatar Grand Prix, in Austria he had eight of his people working on track limits instead of four, and between them he monitored 820 corner passes and sent them to race control. Narrowed down to his 141 reports. Race control elected to delete 51 laps. Thanks to Austria and software improvements, we are able to address those checks and turn them into 150 reports. Here’s a simple example of clicking on a list of reports and selecting “Yes” or “No.”
When it comes to things like track limits, the ROC is currently a data processing facility. Will it evolve in the future? To be honest, I think this could evolve a bit in the direction of VAR. VAR actually appoints someone working remotely to be, in motorsports parlance, a fact judge. But for now, this is a data processing resource designed to assist Race Control.
CB: This reduces the workload on Race Control’s desk and allows us to expand the total amount of work performed by Race Control during a race.The same goes for time penalties.
In the race. Previously, you had to check in Race Control. You are now offline. Through our software tools we can check multiple videos, mark incidents as analyzed and create composite videos for race control and stewards to view. Again, you’re taking a lot of data, filtering that data, and providing a simple package of data to Race Control or the stewards so they can work quickly.
What has changed in Austria?
TM: To summarize the main points about Austria and the present, the first thing to say is that we have changed our nation. Previously, there were three competing data sources in terms of how to identify potential track limits: vehicle detection (which estimates the vehicle’s position relative to the track), loop detection, or manual capture. Detection through someone’s eyes. Essentially, what we concluded after Austria was that all three of these data sources were sending latent reports, and we were able to pull out whatever was being reported incorrectly by the loop and do our analysis. This means that all data correctly reported by the loop will be known if As for whether Loop was also captured by humans, the answer is yes. Therefore, we basically concluded that the loop is not accurate enough. And so far, our most accurate solution has been to have the data analyst watch the video itself. In fact, through loop positioning, his GPS positioning, etc., humans are still winning at this point, so this is an interesting element of the story.
CB: I turned off loops on all circuits unless there were chicanes. Because that just gets in the way of what we’re trying to accomplish. And ultimately, the rule of thumb is that if it’s too close to call, the benefit of the doubt is on the driver.
What’s the next step for ROC’s evolving suite of tools?
TM: What we are trying to do for the future is to improve all our technologies and introduce new technologies. Vehicle positioning continues to be developed to improve accuracy. Next year, he also plans to double the size of ROC from four to eight people and double the connection bandwidth between Trucks and Geneva so more people can work remotely.
CB: The next step is computer vision. This includes shape analysis. There is a line that is the edge of the track, and the software calculates the number of pixels that cross that line.
TM: At the moment, we are “bringing” the situation by saying, “We need to do thousands of checks, but how do we do it?” Because it’s the most accurate solution. What we’re trying to do now is take ROC to the next level, which is to introduce AI software.
Again, as strange as it may sound, this AI-based methodology has a lot to do with current discussions in medicine and the use of computer vision to scan data for cancer screenings, for example. There are similarities. They concluded that they don’t want to use computer vision to diagnose cancer; they want to use computer vision to filter out 80% of cases where cancer clearly isn’t present, and with enough training. This means providing treatment to the doctor who received it. People spend more time paying attention to the 20%. And that’s what we’re aiming for.
So, as I said, 800 is now down to 140 and then down to 50. What we’re aiming for with AI is to take that 800 down to 50 and remove things that clearly don’t need human review. So we currently have two layers of checks, and we add an additional Computer Vision layer upstream. This should allow professional users of ROC to see a smaller number of potential breaches, further reducing the number of reports sent to Race Control and improving overall processing speed.
CB: And that’s going to happen this weekend in Abu Dhabi, where we’re going to run computer vision in parallel. His SBG parent company, the company that provides his RaceWatch platform to the FIA, is called Catapult, the company that wears vests with small receivers on pro athletes and women in the NFL, football, etc. Examples from the NFL include: You can identify every player on the pitch, even in large groups. You can also use that technology for live feeds. It’s the same as the new tool, which allows you to draw a “line of interest.” And the AI learns as it goes.
This year has been a year of tremendous development and improvement. What are the key takeaways from 2023?
TM: I think the important thing is to use technology appropriately and make it work for you. That’s the big thing we accomplished this year. Second, communication is still king. And as we expand, it becomes even more important, and it’s not just in terms of processes and procedures, but relationships with people and how people interact with these new systems, interactions, and Race Control. It was also a learning process in that sense.
Fast forward to year three and our biggest imperative is to expand our facilities and continue to invest in our software. Because that’s how we make great progress. And the final point for me is to embrace new technology and continue to evolve.
I have said repeatedly that at this point in time, humanity is winning in certain areas. While this may be the case now, we feel that ultimately real-time automated police systems are the way forward.