[Help-glpk] branch and bound for MILP - improvement possible?

[HBuesching04]

While reading some articles about exact algorithms for TSP, I stumpled 
about a (possible) improvement for branch and bound at MLP: A delta can 
be calculated to some extent what will happen in the future branches 
without doing all branches. By this you can faster prune nodes.
As a mathematican (now programming with SAP) I was quite pleased with 
myself, but everyone tells me that I must prove the effectness of this 
approach.
As I am quite busy and not very ambitious to earn merits, I am very 
interested, that someone else could do the coding and write some kind 
of official work (diploma...) with it. I think that my idea would 
really fit for such a thing. Is there someone interested in more 
details?
I think that this algorithm might have never been implemented in no LP-
package before.

Best regards
Harald.