Discrete Optimization Model and Algorithm for Driver Planning in Periodic Driver Routing Problem

Table 1

The performance comparison of CPLEX and local branching based method.

Instances	CPLEX		Local Branching		GAPL	TL/TC
ID	ZC	TC	ZL	TL
10_1	138.0	7200	138.0	615	0.00%	0.09
10_2	124.0	7200	118.7	604	-4.27%	0.08
10_3	151.7	7200	148.3	605	-2.24%	0.08
10_4	158.5	7200	148.6	607	-6.25%	0.08
10_5	160.8	7200	122.3	603	-23.90%	0.08
Average					-7.33%	0.08
12_1	171.0	7200	161.3	915	-5.67%	0.13
12_2	165.9	7200	110.9	605	-33.20%	0.08
12_3	187.7	7200	129.5	608	-31.01%	0.08
12_4	225.9	7200	165.0	606	-26.96%	0.08
12_5	223.9	7200	131.4	606	-41.31%	0.08
Average					-27.63%	0.09

Notes: The ID ‘10-1’ means there are 10 customers in this instance, and ‘1’ is the index of the instance. Objective value and CPU time for CPLEX are denoted by Z_C and T_C. Objective value and CPU time for local ranching based method are denoted by Z_L and T_L, and GAP_L = (Z_L - Z_C) / Z_C.