HTL rally results

[devenh]

After completing my first open road rally I thought I would share my experience using HLT.

Harry helped me configure HLT in several pre-race threads:

Rally configuration:
viewtopic.php?f=9&t=1627

iPhone & GPS Time:
viewtopic.php?f=9&t=1627

iPad & XPS150 issues:
viewtopic.php?f=9&t=1621

The rally was the Silver State Challenge Classic (sscc.us), which is a timed 90 mile open road race in Nevada on Highway 318 that is held twice a year in May and September. The 127 competitors select their target speeds and the goal is to finish with exactly the selected average speed. I entered the 100mph class, so my goal was to finish the race in exactly 54:00.00 minutes.

On a timed event it is extremely useful to have a navigator, but I didn't have one. The purpose of the navigator is to tell the driver whether he or she is ahead or behind the target pace.

The attached photos show my car and how I configured the cockpit. iPhone 4s, iPad, XPS150, and a GoPro Hero 3. More on these in a moment.

So I decided to try to configure HLT to be my navigator. Here's what I found.

The main challenge is creating a reference lap that HLT can compare against. The "Rally configuration" link above goes into detail on that. Thanks to Harry's patience with me, I was able to write a spreadsheet that created an XML file that I could export to HLT and use as a reference lap. Anyone who's interested in that can reply to this thread.

The views in HLT contain most of the information one needs while driving the rally, but unfortunately they are not all on the same screen. Not wanting to be distracted by changing screens during the race, I simply opted to run HLT on two devices: an iPhone in Timer view and an iPad (connected to an XPS150) in Speedo view.

The Speedo view showed current speed, max speed, and average speed. The Timer view showed my elapsed time and the all important gap to my reference lap. Basically I would drive the "gap" -- modulating my speed (especially during the last 10 miles) to keep the gap as close to zero as I could.

Each car is started at one minute intervals based on an atomic clock. The key is to start HLT at exactly the right time (see the iPhone vs GPS time thread above for more info). I configured my track with a Standing Go start as I thought this would be more accurate that trying to manually start three devices (iPhone, iPad, and a backup stopwatch).

In pre race practice starts that this process was accurate to about +/- 1 second. Sounds pretty good, but winners often come in within 0.1 seconds of target.

So how could HLT be improved? In priority order:

1. Time based start trigger. This would be a trigger that would start based on GPS time, e.g. 10:04 am. I think this would be a pretty easy addition to HLT.

2. A single screen that displayed lap time, gap, current speed, avg. speed, and max speed. Or, more realistically, a user configurable screen where one can select what elements to display.

3. Time remaining. This would be the difference between the reference lap time and current lap time.

So, how did I do?

Well, I wanted to make sure that my average speed was 100mph or above. I did not want to come in at 99.9 mph! So I was biased to come in below 54:00.00.

According to the iPhone, I finished in 53:59.19. The iPad had me at 53:59.36. My actual time was 53:58.9851. Good enough for fourth place (out of eight in the class).

So I think this is pretty darn close -- iPhone 0.2 seconds off, iPad 0.6 seconds off. Hard to tell whether the difference was due to the Standing Go start or crossing the finish line (finish line GPS coordinates were double checked). In any event, having a time based starting trigger would eliminate any starting variance.

So there you have it. Thanks Harry for your great support!

Deven

[Harry]

Pretty impressive arrangement Thanks for reporting! Although accuracy looks o.k., I'm with you the use case could be improved eliminating the error introduced during start (atom clock + reaction times, vs. GPS / acceleration based time). Which of the two devices has been connected to the XGPS?

- Harry

[devenh]

The iPad was connected to the XPS150.

Would there be any advantage to getting an XPS160? This would allow both devices to share the same GPS source. Would the faster refresh rate improve accuracy?

Also, all the sensor delays were the default settings.

Deven

[Harry]

The XGPS160 seems to allow several connections at once - haven't tested this however... Added accuracy is minimal, it is more about more data points to analyze braking points and things like that. So no "must upgrade" from my point of view.

- Harry

[devenh]

Harry,

If you're up for a challenge, take a look at the race results:

http://www.sscc.us/event_results/results-13-sscc.htm

Many of the class winners were less than 0.1 seconds off!

Deven

[thomas8]

Hi Devenh,

I would be very, very interested in the spreadsheet. Right now I’m in a similar situation as you - I need to drive a known route with a known average speed. I thought this was impossible with HLT, but with a "handcrafted" virtual reference lap it should be possible.
It’s too bad that you need two devices, one to show speed/avg speed and one for the timing, but I think I’ll solve this by using the car’s tachometer (knowing for example that 4th gear at 4200 rpm is ~65.0 km/h) to gauge the correct speed.

I guess my best course of action is to first drive the route and add one waypoint exactly every 250 meters.
From there i can create a virtual reference lap and knowing that the waypoints are 250 meters apart I can create a reference lap by filling in the target time to each respective waypoint.

E.g. if I had to drive 18 km at 64.9 km/h, I would enter a time of 0:00:13,87 for the first waypoint, 0:00:27,73 for the second waypoint and so on.

The .hlptrl format only works with absolute coordinates, right? Or is there a way to use distance relative to the start as a waypoint?

[devenh]

Thomas,

I am happy to share my spreadsheet, but it will probably take quite a bit of explaining. I haven't used it for 8 months and now I am not sure I even understand it

A couple of questions for you: How soon is your road race? How proficient are you with spreadsheets?

I am really busy right now, so you may need to be patient!

You won't have to drive the course. As long as the route is on Google maps, you can get all the GPS coordinates online. The only GPS coordinates you *may* need to get are the starting and finishing lines.

Deven

[thomas8]

Hi Deven,

The first event is next week but I have a few more planned for the rest of the year.
I think I have it mostly figured out without a spreadsheet.

Basically I created a track using start and finish and drove one lap using just cruise control. I finished within 5 seconds of my desired time. Then I edited the resulting .hlptrl file by zeroing the <lineal/> and <lateral/> values and changed the <relativeToStart> values based on the calculated distance and elapsed time where I should be to when driving the average.

So using the 18km/64.9kmh example above, the last fix would be at
<relativeToStart>18000,16:38.46</relativeToStart>,
the second to last fix is at
<relativeToStart>17918,16:33.91</relativeToStart>

and so on. I did this manually for the last 80 fixes (= roughly 2.5 kilometers) before the finish, saved the document and sent it back to my HLT. It imported fine.
Sometime down the line I might attempt to write a python script that fills in the correct times for me based on the distance relative to the start (or finish) and the desired speed. If I ever do this I’ll post the link to github here.

I have not tested it yet, but in theory I should now be able to use this lap as my reference lap which is 100% on time for the last 2.5 km, so if I am able to keep gap at zero, I should arrive right on time.

My only worry is that I recorded the lap using the internal GPS on my iPhone 5, so I’m not sure how accurate it will be. The indicated lap distance was off by 0.8% compared to the official distance, but from comparing the distances/times to the known values from my wheel-based tripmeter on the track, it seems like the error (maybe one or more missed fix(es)?) was before the last 2.5 km so it should be ok.

Already ordered a XGPS160 but it won’t arrive in time for the next event.

It kind of blows my mind that there is no better/easier solution for this on iOS yet.
Many rallies have a class without restrictions for electronic/gps assistants these days and I am convinced that an iOS app with a GPS engine as good as the one in HLT would easily blow devices like the Monit and Terratrip out of the water in terms of value and functionality.
Ideally it would have a big green/red indicator showing if you are too fast or slow (I know the grey/red bar in timer view does the same but it’s a bit small), a configurable text-to-speech countdown (for the finish and for counting down to segment speed changes) and some other useful features.

[gplracerx]

GPS position based timing will never be as accurate or precise as a timing line, unless you're willing to spend the many thousands of dollars necessary to buy high precision GPS mobile and base station units to get centimeter level accuracy. And even then you have the problem of slow update rate. That could be solved too by adding a high precision inertial measuring unit to interpolate between the GPS fixes. I would say north of $10,000 for a system like that. Consumer GPS units like the XGPS150 or the VBOX Sport have a measurement error on the order of 2 meters at best. No matter how precisely you define the start and finish line coordinates, the individual fixes in the car will still have that error.

At a speed of 45 m/s (162 kph or 100 mph ) the timing error will be on the order of plus or minus 0.04 seconds. Using just one fix for the timing line only increases the error by ~40%.

[thomas8]

My XGPS160 arrived sooner than expected so I went out and drove the lap using the virtual track as my reference.

I finished 00m00s 42ms behind my target time, but it seemed like the beep from HLT went off very slightly after I crossed the physical finish line. I think this was more luck than anything else, but i was still quite impressed.

One thing I have noticed is that the difference in recording quality between the built in iPhone GPS and the XGPS160 is immense, as expected from the 1hz vs 10hz and 10m vs 2m accuracy.

Now if only my python was good enough to parse and update the <relativeToStart> times of all ~14000 fixes to match the exact average. I’m not going to update those 14000 fixes by hand.

I reckon that is what the spreadsheet is for, Deven. If you don’t want to share the spreadsheet, is there any chance that you could tell me how you did it? Import the data as a csv into excel, modify the TIME_LAP column (i believe it is in seconds from the start) based on the DISTANCE_KM or DISTANCE_MILE and the desired average speed... but then how do you turn it back into HLT compatible xml file?