For the second season in a row, Claytex has taken a (rather long) field trip up to Scotland. This season we moved to a different site, again in the foothills of the Cairngorms, near Pitlochry. But this time, we were able to organise a site with a lot more of everything; more range, 200m instead of 80m; more range increments, 6 instead of 5; more range object variety, 5 instead of 2 objects; more sensors, 17 instead of 15; more time, 5 months instead of 3, and, most importantly, MORE SNOW!
Setup
This year we had the field for a full 5 months, over last year’s 3 with setup occurring at the start of November 2024, packing up the last week in March 2025. While very similar to last season, the setup was a little more challenging, because of the location, with the site being down a (admittedly stunningly beautiful) 6 mile, 20minute offroad drive. The remote location being to our advantage with the size, altitude and weather variety.
The setup took 3 days of assembly, then a further 2 days of sensor setup and range measurement to ensure that we had all the data we would need to baseline the sensors and recreate the site later. While we wanted snow, unfortunately Scotland seemed very keen to oblige early, with most of the setup days we got a healthy dusting, as shown below.
This year, we had a good range of sensors that were logging data every 5 minutes for the full 5 months:
- 4 cameras
- 1 thermal imaging camera
- 6 lidar
- 6 radar
In addition to the weather monitoring equipment to ensure we know what weather there was that was affecting the sensors:
- Distrometer
- Present weather station
- Basic weather station
One of the range objects (left) and the sensor enclosure (below)
The range objects carried over from last year are the retro reflectors (triangular object on left of range object image above) and the main sign, which other than being a big flat board for lots of surface area, include the 5 colours used in UK road signs and an example of a UK speed limit. The additional calibrated greyscale board was attached to most of the shorter-range signs. Finally, at 200m a large triple wide target was put up, this time with the simulated image of a range rover, as shown below
The slightly late arriving, final addition to the tests was a test dummy that is an ADAS target, a stand in for humans to test the autonomous braking. This was added to have a good baseline comparison to a common physical test, using such a target.
Testing Regimen
The testing regimen was very similar to last season, once the baseline tests have been carried out, leave the sensors and range alone, checking on it regularly and fixing anything that the wind, rain or local wildlife broke.
Other than that, the testing consisted of every sensor logging 30 seconds of data every 5 minutes for the full 5 months.
Results
With 5 months of data for 17 sensors, A LOT of data was generated. Over 90TB! And let’s not forget the additional 40TB of data from last year amounting to a lot of data to go through!
But most importantly, what do we have in terms of weather? With last season being uncharacteristically warm, only yielding a cumulative 16 hours of light, moderate and heavy snow, this year we got 52 hours, along with a lot of hail and rain.
What’s Next?
So, we’ve got 130TB of data, what are we going to do with it? Why go to all the trouble to get it in the first place?
So now comes the laborious task of understanding collating and organising the data. The goal of all of this testing is to understand what happens to all the sensors in a wide variety of weather conditions, therefore we need to organise and understand the data enough so that we can pick out the impact of each weather trait on each sensor. The first step of that journey is finding the patterns in the data, and the radar and lidar engineers have all this data to find those patterns.
Once we know those patterns, and can draw a line between cause and effect, then we need to duplicate that effect. We’re a simulation company, so the next step is to simulate this effect within AVelevate. We are working with rFpro to duplicate the test site virtually, as we did for last season’s site, this will give us the playground to develop, investigate and test all the simulated effects against the measured data in a like-for-like scenario, before applying it to a more complex, real-world scenario.
The end goal is to create a simulated sensor, understand the principals of that validation and validation of that sensor in all weather conditions. Once that is done, we will be shortcutting the testing so we can go from a brand new, customer supplied sensor to validated simulated sensor in as little time as possible.
Final Thoughts
While it’s a shame that we don’t see any need to go up to Scotland again for the foreseeable future, it’s for a good reason, we’ve got a shed load of data, and the testing couldn’t have gone better. While this step was by far and away the most fun one, the next step is the most interesting one, working to get a good grasp of what the data means, with help of our partners in the Sim4CAMSens project.
Written by: David Briant – AV Team Leader
Please get in touch if you have any questions or have got a topic in mind that you would like us to write about. You can submit your questions / topics via: Tech Blog Questions / Topic Suggestion.
