Eloy’s time in the Zenzic CAM Scale-up is progressing well. After Day 2 at UTAC Millbrook Proving Ground, we left in a very comfortable place in terms of technology robustness and a good indication of the service’s efficacy. For Day 3 we want to be able to test the technology more rigorously and firm up our internal approach to measuring the software’s benefit.
Day 3 will be at a new venue for us; AssuredCAV (HORIBA MIRA). We will have a new road layout to add to the artificial intelligence software for the required training. We are also going to increase the number of vehicles to 20, with 10 going clockwise and 10 going anticlockwise. This will mean more interactions and more complicated AI forecasts. For the first time we will not regard all vehicles as equal, giving the option to prioritise to emergency vehicles and cyclists.
Finally, we will attempt to make the road more complex with less visibility for the professional drivers so they will be unable to see other vehicles approaching from blind bends. More vehicles will make it more complex for a human to determine the best coordination actions too.
HORIBA MIRA City Course
Coordination with 20 vehicles
A very cold Eloy team
After Day 2 we started to see our first efficacy readings, with a 10% journey time improvement.
With a relatively low vehicle and lap count it is too early to use these measurements to make a true assessment. We expect a few key actions to increase the software benefit towards our simulation output of 20%:
- The base case lap times will be comparatively higher as cars will meet each other more frequently due to lower visibility on the narrow sections of road. On Day 2, the professional drivers could often see other vehicles on the open track environment we have at testbeds.
- When cars meet each other and cause a stoppage, the time cost for this is likely to be higher due to the higher car density (20 cars vs. 10 previously). Traffic jams and delays are directly correlated to car density.
- The app-on interventions will be clearer due to the change to a banner, and therefore easier to follow
- We will look to increase the vehicle speed on the wider sections to a more realistic level (up from 15mph and towards 30mph). Therefore, the time spent per lap on the narrower sections will be higher and the delays caused on these will be a higher proportion of total lap times. Again, the proportion of narrow road to wider road is a parameter with direct correlation to the intervention benefit – if there are no narrow sections, the software isn’t required.
- A new set of professional drivers will be new to the test. We noted that after Day 2 the drivers got far better at organising themselves around the base case, and wanted to beat the software, due to familiarity with testing.