lanelet2_matching

Library to match objects to lanelets

• Links

  • Repository

• Standard Documents
  • CHANGELOG
  • PACKAGE
  • README

README

Lanelet2 Matching Module

This module provides functions to determine in which lanelet an object is/could be currently located.

Matching functions

• Deterministic matching: Find all lanelets to which an object has less than a certain Euclidean distance.

  • as lanelets have an orientation, every lanelet is considered twice: regular and inverted, but with the same distance

• Probabilistic matching: Compute the squared Mahalanobis distance of the object pose to the lanelet to reason about the probability of that match, as suggested by Petrich et al. (DOI:0.1109/ITSC.2013.6728549)

  • this distance includes the orientation, thus every lanelet is considered twice: regular and inverted, but with different distances

• Accounting for traffic rules: In case you expect traffic rule compliant behavior, you can remove non-compliant matches by providing a traffic rule element for the participant you were matching (such as pedestrian lanelets for vehicles or inverted one way lanelets, which means driving in the wrong direction).

Usage

C++

#include <lanelet2_matching/LaneletMatching.h>

// create objects to match
lanelet::matching::Object2d obj; // deterministic
lanelet::matching::ObjectWithCovariance2d obj2; // considering uncertainty

// retrieve lanelet matches from map
auto detMatches = getDeterministicMatches(laneletMap, obj, 4.); // max distance = 4m
auto probMatches = getProbabilisticMatches(laneletMap, obj2, 4.); // max distance = 4m

// remove non-compliant matches (such as driving in the wrong direction)
auto compliantDetMatches = removeNonRuleCompliantMatches(detMatches, trafficRulesPtr);
auto compliantProbMatches = removeNonRuleCompliantMatches(probMatches, trafficRulesPtr);

have a look at the C++ unittests for more examples

Python

import lanelet2
# Note: in the standalone version of lanelet2_matching, the python module was named
#  "lanelet2_matching". Now it is a submodule of lanelet2: "lanelet2.matching"

# create objects to match
obj = lanelet2.matching.Object2d()

# retrieve lanelet matches from map
matches = lanelet2.matching.getDeterministicMatches(lanelet_map, obj, 4.)  # max distance = 4m

# remove non-compliant matches (such as driving in the wrong direction)
compliant_matches = lanelet2.matching.removeNonRuleCompliantMatches(matches, traffic_rules)

have a look at the python unittests for more examples, also supports uncertainty

License

This package is distributed under the 3-Clause BSD License, see LICENSE.