Optimal Control of Diesel Engines: Numerical Methods, Applications, and Experimental Validation
Table 3
Performance of the NLP solvers on test cycles 3 (top) and 4 (bottom). Only the exact Newton methods are shown for IPOPT and WORHP. The regularisation is chosen individually for all discretisation variants such that smooth control-input trajectories are obtained.
SNOPT
IPOPT
WORHP
0.25
1
0
2,097
666
35 (29.9)
16 (26.4)
24 (54.5)
0.40
3
20
3,924
1,256
37 (114.8)
19 (61.1)
24 (108.1)
0.25
5*
10
10,449
669
24 (114.5)
17 (153.6)
33 (338.9)
3.00
15
10
2,574
822
51 (120.8)
18 (39.0)
28 (69.7)
9.60
47
30
2,547
810
179 (554.6)
200 (544.4)
26 (92.7)
58.0
282
30
2,547
791
200 (5,495.5)
26 (448.2)
24 (433.7)
0.25
1
0
2,205
450
45 (29.7)
28 (46.7)
37 (117.4)
0.40
3
500
4,113
1,199
114 (311.5)
29 (89.0)
49 (196.5)
0.25
5*
30
10,989
617
66 (386.1)
76 (680.2)
33 (405.8)
3.00
15
200
2,709
790
197 (331.8)
25
a(65.2)
41 (104.5)
10.1
50
200
2,709
768
200 (659.9)
36 (92.3)
58 (204.2)
61.0
300
300
2,709
797
200 (6,219.3)
30 (2,493.7)
30 (781.6)
aA segmentation error crashed the optimisation at that iteration.
