It took us five months to solve this problem. The challenge lies in proving that it is optimal.
n100w8_1_2-4-7-9-3-9-2-8: LB=2145 UB=2145
While the upper bound (UB) of 2145 is relatively easy to obtain, demonstrating that there is no better solution is quite difficult. With the standard Branch and Bound method, you could run the algorithm for a million seconds and still not achieve a proof.
The initial gap between the lower bound (LB) and upper bound (UB) is less than 10, which may seem manageable, but it is actually one of the most challenging problems.
Throughout this process, we have learned a lot and gained valuable knowledge. What makes this particular instance unique is the extremely low gap-narrowing gain (a term we coined). We believe this is due to structural symmetry. To address this, we would need to implement symmetry breaking, but we will discuss that another time.
There are now only two remaining unresolved instances for INRC2 8weeks on our side!
n080w814-4-9-9-3-6-0-5
n080w820-4-0-9-1-9-6-2