Allow configurable maximum angle for AngleSteeringSafetyTest #2231
Conversation
|
I need this for #2217 |
|
It's not clear this is the right fix... perhaps we should clip to the max in the CANPacker? That still doesn't fail explicitly though. |
|
While it's possible that there is a better fix, I don't have the knowledge to identify or execute it. There are many other instances in these Python test where the code for the vehicles communicates limits and restrictions back to the base tests in I definitely see the wisdom in there being a mechanism for the packer to throw some sort of exception or error, but I think that even that would just result in the fine tuning of test parameters for those cases where the CAN packer failed, like this one. Another option might be to initiate that range of angles to something smaller then I'm really interested in getting something merged so I can remove my temporary fix from #2217 |
The Subaru LKAS_ANGLE message is limited to a signed, 17-bit integer, which is more than enough for it's maximum steering of around 485 degrees (at least for my Ascent). However, the
test_angle_cmd_when_enabled()defaults to twice the maximum steering angle, which will overflow this value and result in erroneously failed tests.I have added a class variable, and logic that will default the test angle to twice the max angle value when it is not specifically configured. This keeps current behavior for existing tests, but will allow me to constraint he maximum test angle to the limits of my CAN bus messages. It's quite possible that other vehicles will be affected as more angle-based steering systems are supported, too.