binpacking | Bin Packing Playground

Now I want to do something on the branch and bound using SCIP, and begin from the branching rule. While I tracking the branching process I found something I cannot understand. Begining from getting the branching variable candidates using SCIPgetLPBranchCands, I get the SCIP_VAR** lpcands, then I select the first variable to branch using SCIPbranchVar. This branching rule works on each focused node.

Consider the branch decision at the number 1 node (the root node), after executing SCIPbranchVar function, SCIPgetChildren return two child nodes (2 and 3), but the child node number of root node changed, two or more child node appeared. Suppose node selector selected the node of number 2 to branch, SCIPgetSiblings return 3 siblings (3, 4, and 5).

To make it clear that what happened, I print the branching decision path using SCIPprintNodeRootPath, and print the relationship between nodes. The result shows that after I branching at a node on the selected variable, the SCIIP branching at the same node on another variable which I don't know.

I have print these information generated by scipoptsuite-7.0.1/scip/examples/Binpacking, no more branching decision is made. But more child nodes appearing after I replace the ryanfoster rule using the branching on the first variable. After that, I try the create child node branching style for my procedure, extra nodes appeared anyway.

I cannot find out what happened and how to totally control the branching process, it is very important for my work in the future for it determined how to design the branching rule. I have even try to read the source code of SCIP, unfortunately, it's too hard for me to read.

My procedure as follows:

main.cpp