12345qiupeng
/
gtsam
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Moved the pre-sync code back into the presync() function call. This allows the Filter and Smoother to synchronize properly even when no call to smoother.update() has been made.

release/4.3a0
Stephen Williams 2013-04-12 13:05:20 +00:00
parent 34d7742d95
commit f3007f5013
1 changed files with 13 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
gtsam_unstable/nonlinear/ConcurrentBatchSmoother.cpp
View File

@ -107,6 +107,15 @@ void ConcurrentBatchSmoother::presync() {
gttic(presync); gttic(presync);
// TODO: Don't let the length of this code fool you. There is considerable work being done here.
// When we start multi-threading the filter and smoother, it would be ideal if the pre-sync stage
// could happen before the filter is stopped.
// Calculate the marginal on the separator from the smoother factors
if(separatorValues_.size() > 0) {
updateSmootherSummarization();
}
gttoc(presync); gttoc(presync);
} }
@ -373,11 +382,14 @@ void ConcurrentBatchSmoother::updateSmootherSummarization() {
// variables that result from marginalizing out all of the other variables // variables that result from marginalizing out all of the other variables
// These marginal factors will be cached for later transmission to the filter using // These marginal factors will be cached for later transmission to the filter using
// linear container factors // linear container factors
// Note: This assumes the ordering already has the separator variables eliminated last
// Clear out any existing smoother summarized factors // Clear out any existing smoother summarized factors
smootherSummarization_.resize(0); smootherSummarization_.resize(0);
// Reorder the system so that the separator keys are eliminated last
// TODO: This is currently being done twice: here and in 'update'. Fix it.
reorder();
// Create the linear factor graph // Create the linear factor graph
GaussianFactorGraph linearFactorGraph = *factors_.linearize(theta_, ordering_); GaussianFactorGraph linearFactorGraph = *factors_.linearize(theta_, ordering_);