Better management of RPP path pruning when cusp points are detected #4763
Conversation
|
This pull request is in conflict. Could you fix it @andreacelani? |
|
@andreacelani, all pull requests must be targeted towards the |
There was a problem hiding this comment.
Heed the CI bots - we need this in main before we can have it in humble so that all new distributions and current releases can have this update (i.e. jazzy which I assume you'll also want eventually 😉 ). Also on linting and DCO sign offs for quality control.
You can leave this PR open after fixing the DCO and linting issues, but also need a main version of it. Generally I'd recommend closing this PR and opening a main for us to discuss over and fix issues on - then once we merge open a humble version. That way any fixes for main get waterfalled into humble without trying to maintain 2 PRs simultaneously and perhaps miss some details by accident.
I did a good sweep at it and generally look good to me other than missing test coverage / linting issues. I'll do a detailed sweep after the main version of this PR is opened, but generally seems good. I assume you tested this fix?
| { | ||
| auto && theta = fmod(angles + M_PI, 2.0 * M_PI); | ||
| if (theta < 0) | ||
| theta += 360; |
| * @return Shortest distance between angles | ||
| */ | ||
| template<typename F, typename T> | ||
| auto shortest_angular_distance( |
There was a problem hiding this comment.
I believe both of these functions can be used from angles rather than reimplemented https://github.com/ros/angles
| @@ -166,38 +160,6 @@ TEST(RegulatedPurePursuitTest, createCarrotMsg) | |||
| EXPECT_EQ(rtn->point.z, 0.01); | |||
| } | |||
|
|
|||
| TEST(RegulatedPurePursuitTest, findVelocitySignChange) | |||
There was a problem hiding this comment.
make sure to add tests for your new methods (isWithinInversionTolerances prunePlan removePosesAfterFirstInversion findFirstPathInversion)
Basic Info
Description of contribution in a few bullet points
the contribution make RPP controller follows better the path given by planner. There is no skip of path points even with tight maneuvers.
This apply to all models (diff drive, ackermann and omni)
fixed.RPP.webm
Description of documentation updates required from your changes
Added the following parameters. They are the same found in MPPI controller
inversion_xy_toleranceinversion_yaw_tolerancethe parameters are useful to terminate an inversion and start the following the next portion of the path
Future work that may be required in bullet points
original RPP coding style has been follow (one single file cpp file). MPPI has more structured organization with a dedicated path_handler.cpp file
-->
For Maintainers: