TL:DR Was it a success?
Overall, we’re moderately happy with Day 1; we’d give it a score of 7 out of 10. We achieved what we had hoped for but unearthed several new challenges that we will need to overcome. We probably should be more satisfied as testing is meant to push the limits, which we did, but we make some rookie errors that meant we didn’t get some of the results we hoped for
A key area of any efficacy test is the ability to collect data that can be used to show whether the technology works or not. We’re pleased that this worked incredibly well across all 10 vehicles and mobile apps, with GPS coordinates, vehicle bearings, and identifiers all working well. This has allowed us to feed the collected data back into our digital twin and observe what happened on the day.
Session 1 - morning
A vehicle approaching a passing place
We set up cones and home-made road signs at points on the track to mark the narrow stretches and the passing places.
We then briefed the drivers giving them each an iPhone with the Eloy app installed which would display the Stop and Proceed messages as they drove the circuit. It was interesting to note that during the briefing session, some of the drivers were quite adamant that they didn’t think tech could improve a driving experience.
Baseline
The first hour of driving was to get the drivers accustomed with the road layout not only for safety reasons but also to allow us to collect data on circuits without any intervention from our technology.
We had created 5 narrow stretches in different environments on the track so we could see how drivers behaved when they had full visibility, as well as over blind stretches, and windy road segments.
Passing points with clear visibility ran very smoothly without any intervention – drivers had a clear line of sight and would wait until other vehicles were through the narrow sections.
For the tougher sections with blind views, it was chaos every time. Each driver had no ability to see round blind bends, so in a real-life situation, vehicles would have had to brake and reverse, or worse. However, as we didn’t physically close off the single-track stretches, drivers could drive into them and pass each other. This was our first mistake.
The wiggly road segment, despite having good visibility, was also tough for the drivers. A lot more concentration is required due to the amount of turning required, which meant they were less successful at observing other vehicles.
Morning app interventions
Sending the drivers on their way
One interesting outcome was that within 3 laps, the software intervention meant vehicles formed platoons – 2 snakes of vehicles travelling in opposite directions. We had predicted to see this emerge as the AI directing the cars is more likely to allow all cars through a narrow stretch of road in 1 direction first rather than staggering the vehicles, and it was great to see it in action.
The simple, visible narrow road segments had a smooth flow, but given that no intervention was also very smooth, there wasn’t much to improve on here. The difficult wiggly section still proved difficult, so we need to explore if the intervention along with turning manoeuvres is too complex.
The blind sections proved to be the best for our software intervention, with 2 specific sections showing better outcomes than the baseline – although I will discuss that later in the context of the road layout mistake.
Lunch - driver interviews
Between the morning and afternoon sessions, we sat down with the drivers for feedback to find out how they were finding the app and how they would like to see the information displayed.
We were delighted to find that they had all really enjoyed the experience and had changed their minds from what they were thinking before the tests and could now see how tech like this can improve driving on narrow rural roads.
Afternoon testing
Road layout errors
As this was our first track day, we didn’t put a huge amount of preparation into the road layout. Although the narrower stretches of road had signs at either end they weren’t actually narrower, so when vehicles met in the narrow sections, they just freely passed each other.
Whilst we could observe these passing errors and even assign a theoretical time penalty, this misses the point. The time delay on a reverse could be significant and may have an impact on other vehicles due to blocking the road.
Therefore, for the next day of testing we will make the narrow sections wide enough for just one car (more cones please!) and force drivers to reverse as they would in the real world. We expect this to then cause far more traffic congestion and show a clear benefit from the baseline.
