Research Article
Adaptive Coordinated Variable Speed Limit between Highway Mainline and On-Ramp with Deep Reinforcement Learning
Table 4
Performance comparison of DRL-ACVSL, EX-ACVSL, and Do-nothing strategies.
| Scenario | Static demand | Dynamic demand | Strategy | DRL-ACVSL | EX-ACVSL | Do-nothing | DRL-ACVSL | EX-ACVSL | Do-nothing |
| Traffic flow (veh/h) | Avg. | 4,291 | 3,998 | 3,909 | 4,818 | 4,208 | 4,041 | Med. | 4,482 | 4,025 | 3,756 | 4,856 | 4,400 | 3,945 | Std. DV | 517 | 468 | 660 | 610 | 656 | 688 |
| Traffic density (veh/km) | Avg. | 108 | 130 | 168 | 115 | 94 | 156 | Med. | 109 | 134 | 177 | 119 | 90 | 162 | Std. DV | 22 | 27 | 30 | 23 | 30 | 34 |
| Traffic speed (km/h) | Avg. | 41 | 32 | 25 | 44 | 47 | 28 | Med. | 42 | 30 | 22 | 42 | 49 | 25 | Std. DV | 6 | 8 | 9 | 10 | 11 | 11 |
|
|