Community Forum
ANVEL Licensing Options

Acceleration Validation



  • I am currently working on a V&V paper for ANVEL. To test the max acceleration performance that ANVEL can realistically represent I have modeled a McLaren F1 . I've gotten it to be realistic up until 110MPH with a confirmed torque curve, gearbox, drag coefficient, mass, and reasonable shift speeds (shift speeds area variable since the F1 is a manual). Have I missed something or is the ODE limited in its high speed representation? I have included a graph comparing my current build vs Actual Acceleration.0_1537286464946_MF1Accel.PNG



  • @collindavenport,

    There are more parameters at work that may effect your acceleration, such as the AirResistance (property of the vehicle), rolling resistance (property of the surface). You may also want to tinker with drivetrain losses (efficiency of the axle, and torque scale on engine), the impact of which will grow with velocity.

    Regards,
    ANVEL Support



  • I've improved the curve a bit

    Air Resistance
    I have the correct drag coefficient and surface area
    <body dragCoefficient="0.32" name="body" physics="on" surfaceArea="1.79">

    Rolling Resistance
    I currently have friction set to 2.0 but I'm not sure what units that is or how to calculate it.
    Front:
    <contact bouncyness="0.1" fds="0.501" friction="2.0" friction2="0.95"/>
    Rear:
    <contact bouncyness="0.1" fds="0.501" friction="2.0" friction2="0.95"/>

    It's a RWD here is my Axle line. The front wheels have a smaller radius than the rear wheels, but in ANVEL the front wheels are spinning less than the rear wheels which shouldn't be the case. I am testing it on the Large Parking Lot map.

    <axle efficiency="1.0" inertia="0.1" name="rearAxle" output1="rightRear" output2="leftRear" powerFactor="1.0" slip="1" type="limitedSlip">
    <ratio value="2.85"/>
    </axle>

    The torque curve is sourced from a dyno test so I don't want to mess with that. Its a manual as well so there is no transfercase line or torque converter

    0_1537475241706_Capture.PNG



  • @collindavenport,

    I have some more stream of consciousness feedback from our development team. They're supercar enthusiasts, so if you want to send over your model for them to take a look, it might help tweak some parameters in addition what we're suggesting here:

    • "Friction" is the longitudinal coefficient of friction (between the wheel and surface) and thus 2.0 will provide unrealistic levels of grip, but a 1.4 scale may make more sense. Part of the problem is that when doing an actual drag race run like this, you aren't slamming the throttle full open immediately otherwise the tires spin too much. ODE Friction doesn't really lose much grip when spinning the tires compared with Pacejka VTI though, so that may not be much of an issue.

    • The rolling resistance property is found on the ODE Surface data entry for Concrete (or your surface of choice), and adjusting that may help.

    • Try dropping the axle efficiency to 0.98 or lower. Initially you are grip limited, but at high speeds, the power factor will be important.

    • You might be spending too little time in 2nd gear, so you may want to adjust the shift speeds a bit.

    • Your fds values may also be a bit on the high side. I think we normally have much lower values. Your front tires are running slower than the rears due to the power being applied to the rears while the fronts are pretty much locked at the actual vehicle speed since they have no other force accelerating them. If your wheels are always spinning too fast though, you may want to try reducing the throttle a bit.

    Regards,
    ANVEL Support



  • In furthering my study on the RWD I used the Dodge SRT Demon which has the same size on all 4 tires. It still has the same issue with the front wheels spinning less than the rear wheels. I noticed that the Sedan-RWD does not have this issue until I put the Dodge's torque curve in. It would seem that there is too much torque on the Rear Axle and the front wheels "glide" or can't keep traction. Do you have any insight on this?



  • Just saw your post after I posted. I'll give some of those thoughts a shot. I'd be happy to send the vehicle files how do you want me to send them?


  • administrators

    @collindavenport,

    Email to anvel_support@quantumsignal.com would be fine!

    Regards,
    ANVEL Support



  • @collindavenport,

    Thanks for sending over the files! From our perspective, the vehicle definition has some issues. Before we dig into it, are you manually driving this vehicle (with the keyboard or joystick)? If programmatically controlled, can we get the code you are using to test out the vehicle?

    On the vehicle def:

    • The FDS values are way too high, which causes the vehicle to slide around like crazy when doing any sort of steering.

    • One key point to note here is that the testers of these vehicles (generating the acceleration curves) are highly talented drivers that balance wheel spin off the line with throttle application. How the test is performed (is the throttle floored at the start? Is a clutch used to keep starting revs high and then blended in precisely?) will impact the results. You nor I could achieve this same acceleration curve in real-world driving without lots and lots of practice. Furthermore, in an ideal case, tires are warmed up, its on a very sticky bit of asphalt, etc, so we don't necessarily think tuning it to maximum performance with a simple vehicle controller is realistic.

    • It is extremely difficult to avoid spinning out the rear wheels on a car like the Demon. You have to baby the throttle as the torque is can provide in 1st gear is well beyond the traction limits of any tire.

    • Pacejka VTI will be the most accurate for pavement acceleration tests, and getting some more accurate tire data will be key to matching these super car results. Their tires have much higher peak grip than a standard road tire, and are super wide, which influences the slip/grip curves.

    • It looks like you are doing a trick of including a separate vehicle object to represent the center of mass, which slightly decreases processing performance, and may influence accuracy as well. It looks like you also included some debugging physics objects that are meant to represent the current wheel rotation rates. This is most likely actively interfering with the dynamics, and should be completely removed.

    Hope this helps provide some insight...

    Regards,
    ANVEL Support