Getting Ready for the Sparkfun 2014 AVC

Hi All,

We will be competing in the Peloton Class again with our Traxxas Slash 2WD brushless rover chassis in this years AVC.

We have upgraded from an APM2.5 to a 3DR Pixhawk navigation processor and are using twin GPSs.

The GPS selection function is quite simple in that it uses the GPS with the most sats at any given moment which does not necessarily guarantee the best HDOP value, but is great for redundancy in that our tests have indicated that at least one sat will always have a lock at any given time.

We have also perfected an Active Braking function. Active Braking allows us to selectively apply reverse throttle (<1500us) when approaching a turn at a high speed to get the rover decelerated to a turn speed (~3m/s) that will not cause a rover rollover.

We can select the amount of breaking (Braking_Percentage) and the difference between the GPS speed and the requested speed (Speed_Err) at which active braking will occur.

I have been able to approach a 90 deg turn at ~9m/s and have the rover decelerate to 3m/s in less than 5m in order to make the turn without rolling over. Active Braking is also very effective in a chicane like how the barrels are placed on the AVC course.

We have also been evaluating DCM heading estimation vs EKF heading estimation and have found that the EKF function virtually eliminates post turn oversteer and provides exceptionally accurate turns.

We do not plan to use line following, but I might fall back to my twin sonars for obstacle avoidance at low speeds.

However I feel that accurate GPS position knowledge of the barrels, hoop, and the ramp along with course practicing on Friday afternoon and Saturday morning will help to optimize the course waypoints.

I believe that the competition is going to be very tough this year compared to last year.


Tom C ArduRover2 Developer

I’m going to be using a stock ArduRover, with single GPS. You’re going to clobber me!

I do, however, plan to see if I can get the LIDAR-Lite working in time.

Based on my experiences from the 2013 AVC, anything can happen no matter how well equipped your rover is in relation to the other entrants.
At the start of each heat, many of the rovers go wild and either run into each other or do not make the first left hand turn and run into the perimeter fence. On the second heat a competitor’s yellow VW buggy went diagonally across the track and hit my rover and bounced over it into the left side hay bales. Fortunately my rover was not damaged and went on to win that heat at an astounding speed of 2m/s:-).
I hope to run at a speed of at least 6m/s this time and plan to practice on the course on late Friday afternoon and early Saturday morning.
TCIII ArduRover2 Developer

Hi Thomas,

I am planning to build a rover and thus I am interested in your twin GPS setup?
I found (very generally) that when you look at two (or more) GPS in parallel the location may vary no matter what hdop or how many sats. For sure the more sats and the lower the hdop the better but still a simple switch between two different GPS might cause some rather dramatic changes in navigation. Or? A weighing/averaging of the coordinates in relation to the number of sats and hdop might help filtering out potential errors. Have you tested anything like this?

Best regards,

You have some very valid points.
The ArduRover2 Developer Team has been reviewing twin GPS data that I have taken during recent testing for the AVC.
We have found that the position data is very comparable for both GPSs and does not vary enough to create a significant heading/position knowledge problem when switching between the two.
I am using the same uBlox LEA-6 GPS for both the primary and secondary GPSs.
We will probably go to some form of averaging/filtering, as you have suggested, down the road when better GPSs become available, but for now we will use whichever GPS has the higher sat count.
One prime reason for using the twin GPS setup is that from our experience we seldom loose lock on both GPSs at the same time. So this is kind of redundancy insurance at the moment:-)
TCIII ArduRover2 Developer

Hi Thomas,
yes, any redundancy is good! Averaging might also help increasing positional accuracy.
I just received my NavSpark NS-RAW. The idea is to use it as the main (RTK-)GPS and the LEA-6 as a failsave on the Pixhawk. Not sure if it will work and how good the NS-RAW really is. I will report.
Kind regards,


How did the NS-RAW work? I just received mine and am considering entering them into the 2015 AVC.

I am going through the RTK lib right now and will likely visit Sparkfun in the next few weeks or so as I live just outside of Boulder.

Howdy! I was wondering why you chose the Traxxas Slash 2WD in particular over other models. Why this, and not some other thing?

Traxxas 2WD Slash was replaced with a brushless 4WD Slash for the 2014 AVC.

Why switch to 4WD for an asphalt course?
And are you using a 1/10 or a 1/16 model? I’m hoping to fit everything into a 1/16 model for next year.

Went to the 4WD version for a number of reasons:
The brushless motor in the 4WD version is much more powerful than the brushed version in the 2WD version.
The brushless motor/transmission in the 4WD version is in the middle of the chassis providing better weight distribution than the 2WD brushed version which is mounted in the rear.
Since the 4WD front wheels are driven by the brushless motor the 4WD version is able to pull itself around corners better than the 2WD version which is pushed around corners by the rear wheels.
Also there is the additional braking provided by the second set of driven wheels.
I have yet to find an asphalt course that did not have pea gravel or other small rocks and fine dust on its surface.
The Traxxas Slash 4WD chassis is a 1/10 scale chassis.

I stuffed a PX4 FMU/IO, not a Pixhawk, into a Traxxas 1/16 Fiesta 4WD chassis. Good luck getting a Pixhawk into a 1/16 chassis and still get the body on.

I guess my thinking was it’s an advantage to have the motor in the back, away from the electronics, for better noise immunity. But then it’s a brushed motor which I’m guessing is more noisy then a brushless one would be, so there went that advantage. Watching the AVC videos, it looks like most of the RC car/trucks are 1/16 scale, but as you have indicated, there isn’t room for the electronics so a lot of them are topless.

So many decisions to make, thanks for the replies!