Merge pull request #1085 from dstl/dist_hyp_meas

Make passing measurements in distance hypothesiser optional

stonesoup/hypothesiser/distance.py

@@ -26,6 +26,10 @@ class DistanceHypothesiser(Hypothesiser):
     include_all: bool = Property(
         default=False,
         doc="If `True`, hypotheses beyond missed distance will be returned. Default `False`")
+    predict_with_measurements: bool = Property(
+        default=False,
+        doc="Whether to pass measurement/detection to the predictor. Default `False`"
+    )
 
     def hypothesise(self, track, detections, timestamp, **kwargs):
         """ Evaluate and return all track association hypotheses.
@@ -68,8 +72,12 @@ def hypothesise(self, track, detections, timestamp, **kwargs):
         for detection in detections:
 
             # Re-evaluate prediction
-            prediction = self.predictor.predict(
-                track, timestamp=detection.timestamp, measurement=detection, **kwargs)
+            if self.predict_with_measurements:
+                prediction = self.predictor.predict(
+                    track, timestamp=detection.timestamp, measurement=detection, **kwargs)
+            else:
+                prediction = self.predictor.predict(
+                    track, timestamp=detection.timestamp, **kwargs)
 
             # Compute measurement prediction and distance measure
             measurement_prediction = self.updater.predict_measurement(

stonesoup/hypothesiser/tests/test_distance.py

@@ -2,6 +2,7 @@
 import datetime
 
 import numpy as np
+import pytest
 
 from ..distance import DistanceHypothesiser
 from ...types.detection import Detection
@@ -10,7 +11,8 @@
 from ... import measures
 
 
-def test_mahalanobis(predictor, updater):
+@pytest.mark.parametrize('predict_with_measurements', [True, False])
+def test_mahalanobis(predictor, updater, predict_with_measurements):
 
     timestamp = datetime.datetime.now()
     track = Track([GaussianState(np.array([[0]]), np.array([[1]]), timestamp)])
@@ -21,7 +23,8 @@ def test_mahalanobis(predictor, updater):
 
     measure = measures.Mahalanobis()
     hypothesiser = DistanceHypothesiser(
-        predictor, updater, measure=measure, missed_distance=3)
+        predictor, updater, measure=measure, missed_distance=3,
+        predict_with_measurements=predict_with_measurements)
 
     hypotheses = hypothesiser.hypothesise(track, detections, timestamp)
 

stonesoup/proposal/base.py

@@ -7,7 +7,9 @@ class Proposal(Base):
     @abstractmethod
     def rvs(self, *args, **kwargs):
         r"""Proposal noise/sample generation function
+
         Generates samples from the proposal.
+
         Parameters
         ----------
         state: :class:`~.State`

stonesoup/proposal/simple.py

@@ -27,14 +27,21 @@ class PriorAsProposal(Proposal):
     def rvs(self, prior: State, measurement=None, time_interval=None,
             **kwargs) -> Union[StateVector, StateVectors]:
         """Generate samples from the proposal.
+
         Parameters
         ----------
         state: :class:`~.State`
             The state to generate samples from.
+        measurement: :class:`~.Detection`
+            the measurement that will preferably used to get time interval
+            if provided(the default is `None`)
+        time_interval: :class:`datetime.time_delta`
+            time interval of the prediction is needed to propagate the states
+
         Returns
         -------
-        : :class:`~.ParticlePrediction` with samples drawn from the updated proposal
-
+        : :class:`~.ParticlePrediction`
+            State with samples drawn from the updated proposal
         """
 
         if measurement is not None:
@@ -55,7 +62,7 @@ def rvs(self, prior: State, measurement=None, time_interval=None,
 
 
 class KFasProposal(Proposal):
-    """This proposal uses the kalman filter prediction and update steps to
+    """This proposal uses the Kalman filter prediction and update steps to
     generate new set of particles and weights
     """
     predictor: Predictor = Property(
@@ -66,20 +73,23 @@ class KFasProposal(Proposal):
     def rvs(self, prior: State, measurement: Detection = None, time_interval=None,
             **kwargs):
         """Generate samples from the proposal.
-            Use the kalman filter predictor and updater to create a new distribution
+
+        Use the Kalman filter predictor and updater to create a new distribution
+
         Parameters
         ----------
         state: :class:`~.State`
             The state to generate samples from.
         measurement: :class:`~.Detection`
-            the measurement that is used in the update step of the kalman prediction,
+            the measurement that is used in the update step of the Kalman prediction,
             (the default is `None`)
         time_interval: :class:`datetime.time_delta`
             time interval of the prediction is needed to propagate the states
 
         Returns
         -------
         : :class:`~.ParticlePrediction`
+            State with samples drawn from the updated proposal
         """
 
         if measurement is not None: