Open-Harbor/osrm-backend
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 3 Wiki Activity
9231335eef
osrm-backend/plugins/match.hpp
Daniel J. Hofmann 9231335eef Use Intel TBB's parallel_sort even for nested parallelism. 
TBB has a global task scheduler (that's one of the reason TBB is not
linked statically but dyanmically instead). This allows control over all
running threads, enabling us to use nested parallelism and the scheduler
doing all the task allocation itself.

That is, nested parallel execution such as in

    parallel_for(seq, [](const auto& rng){
      parallel_sort(rng);
    });

is no problem at all, as the scheduler still claims control over the
global environment.

Therefore, use `parallel_sort` Range overload where possible.

References:

- https://www.threadingbuildingblocks.org/docs/help/hh_goto.htm#reference/algorithms.htm
- https://www.threadingbuildingblocks.org/docs/help/hh_goto.htm#reference/algorithms/parallel_sort_func.htm
- https://www.threadingbuildingblocks.org/docs/help/hh_goto.htm#reference/task_scheduler.htm
- https://www.threadingbuildingblocks.org/docs/help/hh_goto.htm#reference/task_scheduler/task_scheduler_init_cls.htm
- https://www.threadingbuildingblocks.org/docs/help/hh_goto.htm#tbb_userguide/Initializing_and_Terminating_the_Library.htm
2015-09-28 20:26:03 +02:00

383 lines
15 KiB
C++
Raw Blame History

/*
Copyright (c) 2015, Project OSRM contributors
All rights reserved.
Redistribution and use in source and binary forms, with or without modification,
are permitted provided that the following conditions are met:
Redistributions of source code must retain the above copyright notice, this list
of conditions and the following disclaimer.
Redistributions in binary form must reproduce the above copyright notice, this
list of conditions and the following disclaimer in the documentation and/or
other materials provided with the distribution.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR
ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#ifndef MATCH_HPP
#define MATCH_HPP
#include "plugin_base.hpp"
#include "../algorithms/bayes_classifier.hpp"
#include "../algorithms/object_encoder.hpp"
#include "../data_structures/search_engine.hpp"
#include "../descriptors/descriptor_base.hpp"
#include "../descriptors/json_descriptor.hpp"
#include "../routing_algorithms/map_matching.hpp"
#include "../util/compute_angle.hpp"
#include "../util/integer_range.hpp"
#include "../util/json_logger.hpp"
#include "../util/json_util.hpp"
#include "../util/string_util.hpp"
#include <tbb/parallel_sort.h>
#include <cstdlib>
#include <algorithm>
#include <memory>
#include <string>
#include <vector>
template <class DataFacadeT> class MapMatchingPlugin : public BasePlugin
{
constexpr static const unsigned max_number_of_candidates = 10;
std::shared_ptr<SearchEngine<DataFacadeT>> search_engine_ptr;
using ClassifierT = BayesClassifier<LaplaceDistribution, LaplaceDistribution, double>;
using TraceClassification = ClassifierT::ClassificationT;
public:
MapMatchingPlugin(DataFacadeT *facade, const int max_locations_map_matching)
: descriptor_string("match"), facade(facade),
max_locations_map_matching(max_locations_map_matching),
// the values where derived from fitting a laplace distribution
// to the values of manually classified traces
classifier(LaplaceDistribution(0.005986, 0.016646),
LaplaceDistribution(0.054385, 0.458432),
0.696774) // valid apriori probability
{
search_engine_ptr = std::make_shared<SearchEngine<DataFacadeT>>(facade);
}
virtual ~MapMatchingPlugin() {}
const std::string GetDescriptor() const final override { return descriptor_string; }
TraceClassification
classify(const float trace_length, const float matched_length, const int removed_points) const
{
const double distance_feature = -std::log(trace_length) + std::log(matched_length);
// matched to the same point
if (!std::isfinite(distance_feature))
{
return std::make_pair(ClassifierT::ClassLabel::NEGATIVE, 1.0);
}
const auto label_with_confidence = classifier.classify(distance_feature);
return label_with_confidence;
}
bool getCandiates(const std::vector<FixedPointCoordinate> &input_coords,
const double gps_precision,
std::vector<double> &sub_trace_lengths,
osrm::matching::CandidateLists &candidates_lists)
{
double query_radius = 10 * gps_precision;
double last_distance =
coordinate_calculation::great_circle_distance(input_coords[0], input_coords[1]);
sub_trace_lengths.resize(input_coords.size());
sub_trace_lengths[0] = 0;
for (const auto current_coordinate : osrm::irange<std::size_t>(0, input_coords.size()))
{
bool allow_uturn = false;
if (0 < current_coordinate)
{
last_distance = coordinate_calculation::great_circle_distance(
input_coords[current_coordinate - 1], input_coords[current_coordinate]);
sub_trace_lengths[current_coordinate] +=
sub_trace_lengths[current_coordinate - 1] + last_distance;
}
if (input_coords.size() - 1 > current_coordinate && 0 < current_coordinate)
{
double turn_angle = ComputeAngle::OfThreeFixedPointCoordinates(
input_coords[current_coordinate - 1], input_coords[current_coordinate],
input_coords[current_coordinate + 1]);
// sharp turns indicate a possible uturn
if (turn_angle <= 90.0 || turn_angle >= 270.0)
{
allow_uturn = true;
}
}
std::vector<std::pair<PhantomNode, double>> candidates;
facade->IncrementalFindPhantomNodeForCoordinateWithMaxDistance(
input_coords[current_coordinate], candidates, query_radius);
// sort by foward id, then by reverse id and then by distance
tbb::parallel_sort(
candidates, [](const std::pair<PhantomNode, double> &lhs,
const std::pair<PhantomNode, double> &rhs)
{
return lhs.first.forward_node_id < rhs.first.forward_node_id ||
(lhs.first.forward_node_id == rhs.first.forward_node_id &&
(lhs.first.reverse_node_id < rhs.first.reverse_node_id ||
(lhs.first.reverse_node_id == rhs.first.reverse_node_id &&
lhs.second < rhs.second)));
});
auto new_end = std::unique(
candidates.begin(), candidates.end(), [](const std::pair<PhantomNode, double> &lhs,
const std::pair<PhantomNode, double> &rhs)
{
return lhs.first.forward_node_id == rhs.first.forward_node_id &&
lhs.first.reverse_node_id == rhs.first.reverse_node_id;
});
candidates.resize(new_end - candidates.begin());
if (!allow_uturn)
{
const auto compact_size = candidates.size();
for (const auto i : osrm::irange<std::size_t>(0, compact_size))
{
// Split edge if it is bidirectional and append reverse direction to end of list
if (candidates[i].first.forward_node_id != SPECIAL_NODEID &&
candidates[i].first.reverse_node_id != SPECIAL_NODEID)
{
PhantomNode reverse_node(candidates[i].first);
reverse_node.forward_node_id = SPECIAL_NODEID;
candidates.push_back(std::make_pair(reverse_node, candidates[i].second));
candidates[i].first.reverse_node_id = SPECIAL_NODEID;
}
}
}
// sort by distance to make pruning effective
tbb::parallel_sort(candidates, [](const std::pair<PhantomNode, double> &lhs,
const std::pair<PhantomNode, double> &rhs)
{
return lhs.second < rhs.second;
});
candidates_lists.push_back(std::move(candidates));
}
return true;
}
osrm::json::Object submatchingToJSON(const osrm::matching::SubMatching &sub,
const RouteParameters &route_parameters,
const InternalRouteResult &raw_route)
{
osrm::json::Object subtrace;
if (route_parameters.classify)
{
subtrace.values["confidence"] = sub.confidence;
}
JSONDescriptor<DataFacadeT> json_descriptor(facade);
json_descriptor.SetConfig(route_parameters);
subtrace.values["hint_data"] = json_descriptor.BuildHintData(raw_route);
if (route_parameters.geometry || route_parameters.print_instructions)
{
DescriptionFactory factory;
FixedPointCoordinate current_coordinate;
factory.SetStartSegment(raw_route.segment_end_coordinates.front().source_phantom,
raw_route.source_traversed_in_reverse.front());
for (const auto i :
osrm::irange<std::size_t>(0, raw_route.unpacked_path_segments.size()))
{
for (const PathData &path_data : raw_route.unpacked_path_segments[i])
{
current_coordinate = facade->GetCoordinateOfNode(path_data.node);
factory.AppendSegment(current_coordinate, path_data);
}
factory.SetEndSegment(raw_route.segment_end_coordinates[i].target_phantom,
raw_route.target_traversed_in_reverse[i],
raw_route.is_via_leg(i));
}
factory.Run(route_parameters.zoom_level);
// we need because we don't run path simplification
for (auto &segment : factory.path_description)
{
segment.necessary = true;
}
if (route_parameters.geometry)
{
subtrace.values["geometry"] = factory.AppendGeometryString(route_parameters.compression);
}
if (route_parameters.print_instructions)
{
std::vector<typename JSONDescriptor<DataFacadeT>::Segment> temp_segments;
subtrace.values["instructions"] = json_descriptor.BuildTextualDescription(factory, temp_segments);
}
}
subtrace.values["indices"] = osrm::json::make_array(sub.indices);
osrm::json::Array points;
for (const auto &node : sub.nodes)
{
points.values.emplace_back(
osrm::json::make_array(node.location.lat / COORDINATE_PRECISION,
node.location.lon / COORDINATE_PRECISION));
}
subtrace.values["matched_points"] = points;
osrm::json::Array names;
for (const auto &node : sub.nodes)
{
names.values.emplace_back( facade->get_name_for_id(node.name_id) );
}
subtrace.values["matched_names"] = names;
return subtrace;
}
int HandleRequest(const RouteParameters &route_parameters,
osrm::json::Object &json_result) final override
{
// check number of parameters
if (!check_all_coordinates(route_parameters.coordinates))
{
json_result.values["status"] = "Invalid coordinates.";
return 400;
}
std::vector<double> sub_trace_lengths;
osrm::matching::CandidateLists candidates_lists;
const auto &input_coords = route_parameters.coordinates;
const auto &input_timestamps = route_parameters.timestamps;
if (input_timestamps.size() > 0 && input_coords.size() != input_timestamps.size())
{
json_result.values["status"] =
"Number of timestamps does not match number of coordinates .";
return 400;
}
// enforce maximum number of locations for performance reasons
if (max_locations_map_matching > 0 &&
static_cast<int>(input_coords.size()) < max_locations_map_matching)
{
json_result.values["status"] = "Too many coodindates.";
return 400;
}
// enforce maximum number of locations for performance reasons
if (static_cast<int>(input_coords.size()) < 2)
{
json_result.values["status"] = "At least two coordinates needed.";
return 400;
}
const bool found_candidates = getCandiates(input_coords, route_parameters.gps_precision,
sub_trace_lengths, candidates_lists);
if (!found_candidates)
{
json_result.values["status"] = "No suitable matching candidates found.";
return 400;
}
// setup logging if enabled
if (osrm::json::Logger::get())
osrm::json::Logger::get()->initialize("matching");
// call the actual map matching
osrm::matching::SubMatchingList sub_matchings;
search_engine_ptr->map_matching(candidates_lists, input_coords, input_timestamps,
route_parameters.matching_beta,
route_parameters.gps_precision, sub_matchings);
if (sub_matchings.empty())
{
json_result.values["status"] = "No matchings found.";
return 400;
}
osrm::json::Array matchings;
for (auto &sub : sub_matchings)
{
// classify result
if (route_parameters.classify)
{
double trace_length =
sub_trace_lengths[sub.indices.back()] - sub_trace_lengths[sub.indices.front()];
TraceClassification classification =
classify(trace_length, sub.length,
(sub.indices.back() - sub.indices.front() + 1) - sub.nodes.size());
if (classification.first == ClassifierT::ClassLabel::POSITIVE)
{
sub.confidence = classification.second;
}
else
{
sub.confidence = 1 - classification.second;
}
}
BOOST_ASSERT(sub.nodes.size() > 1);
// FIXME we only run this to obtain the geometry
// The clean way would be to get this directly from the map matching plugin
InternalRouteResult raw_route;
PhantomNodes current_phantom_node_pair;
for (unsigned i = 0; i < sub.nodes.size() - 1; ++i)
{
current_phantom_node_pair.source_phantom = sub.nodes[i];
current_phantom_node_pair.target_phantom = sub.nodes[i + 1];
raw_route.segment_end_coordinates.emplace_back(current_phantom_node_pair);
}
search_engine_ptr->shortest_path(
raw_route.segment_end_coordinates,
std::vector<bool>(raw_route.segment_end_coordinates.size(), true), raw_route);
BOOST_ASSERT(raw_route.shortest_path_length != INVALID_EDGE_WEIGHT);
matchings.values.emplace_back(submatchingToJSON(sub, route_parameters, raw_route));
}
if (osrm::json::Logger::get())
osrm::json::Logger::get()->render("matching", json_result);
json_result.values["matchings"] = matchings;
return 200;
}
private:
std::string descriptor_string;
DataFacadeT *facade;
int max_locations_map_matching;
ClassifierT classifier;
};
#endif // MATCH_HPP
Reference in New Issue View Git Blame Copy Permalink