Discuss Findings

best flow gain +0.50942 +/- 0.00638 overall; mean mid-cluster gain +0.24317 +/- 0.00275 overall; reconstructed from 3 sampled sanitized outputs x 3 trial seeds, density 0.30 with anticipation 0.35 gives flow gains of +0.22581 +/- 0.02615 (cluster 0.0), +0.21237 +/- 0.03459 (cluster 0.55), and +0.19409 +/- 0.03041 (cluster 0.9), while anticipation 0.70 gives +0.50860 +/- 0.03945, +0.51935 +/- 0.03597, and +0.52437 +/- 0.03441 with corresponding jam fractions 0.23391 +/- 0.02805, 0.22912 +/- 0.02346, and 0.27025 +/- 0.02385; the best-condition summaries choose density 0.30 and anticipation 0.70 in every sampled output, with cluster strength 0.9 most often

Anticipation reliably improves throughput in incident-rich traffic, and the best summaries overwhelmingly favor the densest tested road together with the highest anticipation fraction. In the reconstructed condition grid, moving from 35% to 70% anticipatory drivers roughly doubles the flow gain while sharply reducing jam fractions. The advantage remains positive from unclustered to strongly clustered incidents, so the effect is not confined to one persistence regime. The likely mechanism is earlier braking and gap management near incident zones, which damps stop-and-go wave growth before localized disruptions spread into network-scale jams.