paramsd: fix unbounded yaw rate while calibrating (#34806)

* rm

* comments

* default values when yaw rate invalid

* clean up

Revert "default values when yaw rate invalid"

This reverts commit e983abb3b0.

do the same for yaw rate we do for roll

and

* 1 is fine

* update refs

---------

Co-authored-by: Kacper Rączy <gfw.kra@gmail.com>

selfdrive/locationd/locationd.py

@@ -147,13 +147,13 @@ class LocationEstimator:
       self.car_speed = abs(msg.vEgo)
 
     elif which == "liveCalibration":
+      # Note that we use this message during calibration
       if len(msg.rpyCalib) > 0:
         calib = np.array(msg.rpyCalib)
         if calib.min() < -CALIB_RPY_SANITY_CHECK or calib.max() > CALIB_RPY_SANITY_CHECK:
           return HandleLogResult.INPUT_INVALID
 
         self.device_from_calib = rot_from_euler(calib)
-        self.calibrated = msg.calStatus == log.LiveCalibrationData.Status.calibrated
 
     elif which == "cameraOdometry":
       if not self._validate_timestamp(t):

selfdrive/locationd/paramsd.py

@@ -46,18 +46,25 @@ class ParamsLearner:
     self.yaw_rate_std = 0.0
     self.roll = 0.0
     self.steering_angle = 0.0
-    self.roll_valid = False
 
   def handle_log(self, t, which, msg):
     if which == 'livePose':
       device_pose = Pose.from_live_pose(msg)
       calibrated_pose = self.calibrator.build_calibrated_pose(device_pose)
+
+      yaw_rate_valid = msg.angularVelocityDevice.valid and self.calibrator.calib_valid
+      yaw_rate_valid = yaw_rate_valid and 0 < self.yaw_rate_std < 10  # rad/s
+      yaw_rate_valid = yaw_rate_valid and abs(self.yaw_rate) < 1  # rad/s
+      if yaw_rate_valid:
         self.yaw_rate, self.yaw_rate_std = calibrated_pose.angular_velocity.z, calibrated_pose.angular_velocity.z_std
+      else:
+        # This is done to bound the yaw rate estimate when localizer values are invalid or calibrating
+        self.yaw_rate, self.yaw_rate_std = 0.0, np.radians(1)
 
       localizer_roll, localizer_roll_std = device_pose.orientation.x, device_pose.orientation.x_std
       localizer_roll_std = np.radians(1) if np.isnan(localizer_roll_std) else localizer_roll_std
-      self.roll_valid = (localizer_roll_std < ROLL_STD_MAX) and (ROLL_MIN < localizer_roll < ROLL_MAX) and msg.sensorsOK
+      roll_valid = (localizer_roll_std < ROLL_STD_MAX) and (ROLL_MIN < localizer_roll < ROLL_MAX) and msg.sensorsOK
-      if self.roll_valid:
+      if roll_valid:
         roll = localizer_roll
         # Experimentally found multiplier of 2 to be best trade-off between stability and accuracy or similar?
         roll_std = 2 * localizer_roll_std
@@ -67,14 +74,8 @@ class ParamsLearner:
         roll_std = np.radians(10.0)
       self.roll = np.clip(roll, self.roll - ROLL_MAX_DELTA, self.roll + ROLL_MAX_DELTA)
 
-      yaw_rate_valid = msg.angularVelocityDevice.valid and self.calibrator.calib_valid
-      yaw_rate_valid = yaw_rate_valid and 0 < self.yaw_rate_std < 10  # rad/s
-      yaw_rate_valid = yaw_rate_valid and abs(self.yaw_rate) < 1  # rad/s
-
       if self.active:
         if msg.posenetOK:
-
-          if yaw_rate_valid:
           self.kf.predict_and_observe(t,
                                       ObservationKind.ROAD_FRAME_YAW_RATE,
                                       np.array([[-self.yaw_rate]]),

selfdrive/test/process_replay/ref_commit

@@ -1 +1 @@
-87088a3540ddfc26356d9b71a3a4f40efcbc9b3b
+465979047295dad5da3a552db9227ed776a26a79