Refactor calling optimization into DispatchOptimization. (#729)

I noticed that @jihoonl opened the PR #726 which performs a similar thing. As discussed in googlecartographer/cartographer_ros#613 (@cschuet  has already taken a look), I pulled this out of #481 (a really old PR whose merging has been postponed), which is an example where re-running optimization is triggered from elsewhere as well (besides from `ComputeConstraintsForNode`). This refactoring makes libcartographer friendlier for use cases such as that one.

An important detail is that I have changed the condition in `WaitForAllComputations` to also check if the work queue is empty. If there are other things on the worker queue besides `ComputeConstraintsForNode`, currently we will wrongfully conclude that all computations are done. (This detail was merged in #754, so it's no longer in the diff of this PR).

Also missing is the same thing for 3D. I can add that when we settle on this.

Also, I suggest that `run_loop_closure` gets renamed to `run_optimization`.
master
Juraj Oršulić 2018-01-05 10:43:56 +01:00 committed by Wally B. Feed
parent 923d643b86
commit 58d94aaa68
4 changed files with 59 additions and 32 deletions

View File

@ -269,18 +269,21 @@ void PoseGraph::ComputeConstraintsForNode(
}
constraint_builder_.NotifyEndOfNode();
++num_nodes_since_last_loop_closure_;
CHECK(!run_loop_closure_);
if (options_.optimize_every_n_nodes() > 0 &&
num_nodes_since_last_loop_closure_ > options_.optimize_every_n_nodes()) {
CHECK(!run_loop_closure_);
run_loop_closure_ = true;
// If there is a 'work_queue_' already, some other thread will take care.
if (work_queue_ == nullptr) {
work_queue_ = common::make_unique<std::deque<std::function<void()>>>();
HandleWorkQueue();
}
DispatchOptimization();
}
}
void PoseGraph::DispatchOptimization() {
run_loop_closure_ = true;
// If there is a 'work_queue_' already, some other thread will take care.
if (work_queue_ == nullptr) {
work_queue_ = common::make_unique<std::deque<std::function<void()>>>();
HandleWorkQueue();
}
}
common::Time PoseGraph::GetLatestNodeTime(
const mapping::NodeId& node_id, const mapping::SubmapId& submap_id) const {
common::Time time = trajectory_nodes_.at(node_id).constant_data->time;
@ -384,8 +387,8 @@ void PoseGraph::FinishTrajectory(const int trajectory_id) {
for (const auto& submap : submap_data_.trajectory(trajectory_id)) {
submap_data_.at(submap.id).state = SubmapState::kFinished;
}
// TODO(jihoonl): Refactor HandleWorkQueue() logic from
// ComputeConstraintsForNode and call from here
CHECK(!run_loop_closure_);
DispatchOptimization();
});
}
@ -505,14 +508,21 @@ void PoseGraph::AddTrimmer(std::unique_ptr<mapping::PoseGraphTrimmer> trimmer) {
}
void PoseGraph::RunFinalOptimization() {
{
common::MutexLocker locker(&mutex_);
AddWorkItem([this]() REQUIRES(mutex_) {
optimization_problem_.SetMaxNumIterations(
options_.max_num_final_iterations());
DispatchOptimization();
});
AddWorkItem([this]() REQUIRES(mutex_) {
optimization_problem_.SetMaxNumIterations(
options_.optimization_problem_options()
.ceres_solver_options()
.max_num_iterations());
});
}
WaitForAllComputations();
optimization_problem_.SetMaxNumIterations(
options_.max_num_final_iterations());
RunOptimization();
optimization_problem_.SetMaxNumIterations(
options_.optimization_problem_options()
.ceres_solver_options()
.max_num_iterations());
}
void PoseGraph::RunOptimization() {

View File

@ -215,6 +215,9 @@ class PoseGraph : public mapping::PoseGraph {
// Whether the optimization has to be run before more data is added.
bool run_loop_closure_ GUARDED_BY(mutex_) = false;
// Schedules optimization (i.e. loop closure) to run.
void DispatchOptimization() REQUIRES(mutex_);
// Current optimization problem.
pose_graph::OptimizationProblem optimization_problem_;
pose_graph::ConstraintBuilder constraint_builder_ GUARDED_BY(mutex_);

View File

@ -286,15 +286,19 @@ void PoseGraph::ComputeConstraintsForNode(
}
constraint_builder_.NotifyEndOfNode();
++num_nodes_since_last_loop_closure_;
CHECK(!run_loop_closure_);
if (options_.optimize_every_n_nodes() > 0 &&
num_nodes_since_last_loop_closure_ > options_.optimize_every_n_nodes()) {
CHECK(!run_loop_closure_);
run_loop_closure_ = true;
// If there is a 'work_queue_' already, some other thread will take care.
if (work_queue_ == nullptr) {
work_queue_ = common::make_unique<std::deque<std::function<void()>>>();
HandleWorkQueue();
}
DispatchOptimization();
}
}
void PoseGraph::DispatchOptimization() {
run_loop_closure_ = true;
// If there is a 'work_queue_' already, some other thread will take care.
if (work_queue_ == nullptr) {
work_queue_ = common::make_unique<std::deque<std::function<void()>>>();
HandleWorkQueue();
}
}
@ -401,8 +405,8 @@ void PoseGraph::FinishTrajectory(const int trajectory_id) {
for (const auto& submap : submap_data_.trajectory(trajectory_id)) {
submap_data_.at(submap.id).state = SubmapState::kFinished;
}
// TODO(jihoonl): Refactor HandleWorkQueue() logic from
// ComputeConstraintsForNode and call from here
CHECK(!run_loop_closure_);
DispatchOptimization();
});
}
@ -507,14 +511,21 @@ void PoseGraph::AddTrimmer(std::unique_ptr<mapping::PoseGraphTrimmer> trimmer) {
}
void PoseGraph::RunFinalOptimization() {
{
common::MutexLocker locker(&mutex_);
AddWorkItem([this]() REQUIRES(mutex_) {
optimization_problem_.SetMaxNumIterations(
options_.max_num_final_iterations());
DispatchOptimization();
});
AddWorkItem([this]() REQUIRES(mutex_) {
optimization_problem_.SetMaxNumIterations(
options_.optimization_problem_options()
.ceres_solver_options()
.max_num_iterations());
});
}
WaitForAllComputations();
optimization_problem_.SetMaxNumIterations(
options_.max_num_final_iterations());
RunOptimization();
optimization_problem_.SetMaxNumIterations(
options_.optimization_problem_options()
.ceres_solver_options()
.max_num_iterations());
}
void PoseGraph::LogResidualHistograms() {

View File

@ -219,6 +219,9 @@ class PoseGraph : public mapping::PoseGraph {
// Whether the optimization has to be run before more data is added.
bool run_loop_closure_ GUARDED_BY(mutex_) = false;
// Schedules optimization (i.e. loop closure) to run.
void DispatchOptimization() REQUIRES(mutex_);
// Current optimization problem.
pose_graph::OptimizationProblem optimization_problem_;
pose_graph::ConstraintBuilder constraint_builder_ GUARDED_BY(mutex_);