A Simulated Annealing Approach for the Train Design Optimization Problem

Table 7

Performance comparison of our SA approach and other methods; computing times are given in seconds. The bottom row shows a lower bound on the optimal cost, computed as explained in Section 5.3. (all figures rounded to integers).

Number of	Example		Data Set 1		Data Set 2
Nodes	4		94		221
Blocks	5		239		369
Arcs	5		134		294
Crew Segments	5		154		154
	cost	time	cost	time	cost	time
[21]	—		2,043,471	45	3,187,999	225
[20]	—		2,032,516	2,201	3,188,500	2,067
[18]	—		2,018,644	80,000	3,152,019	75,000
[23]	14,916	0	2,069,752	342	3,240,177	2,700
SA	14,916	1	2,005,514	181	3,170,815	1,991
Lower bound	10,956		1,660,361		2,751,213