Smoothing the wave doesn't dissolve the bottleneck.
Speed Harmonization tries to dampen stop-and-go shockwaves at freeway weaving bottlenecks by recommending lower speeds upstream — giving downstream merging traffic room to clear. A revised VSL-VSA algorithm drives the recommendations from a blend of detector and BSM data.
Each speed-harmonization zone is split into a downstream data-collection segment (2,000 ft past the merging gore) and an upstream broadcasting segment (2,000 ft before it). The edge unit runs a revised VSL-VSA algorithm and pushes recommended speeds via RSMs to approaching CVs.
Loop detectors at upstream and downstream VDS positions provide 5-minute occupancy and speed. CVs in the downstream segment broadcast BSMs at 10 Hz with their speeds.
Speed input to the algorithm is v_m = (1−α)·v_det + α·v_BSM with α = 0.1. Detector data anchors reliability (100% observation rate); BSM data captures dynamics.
If detector occupancy exceeds 20% (the user-defined threshold), the algorithm activates. Below that, no speed limit is broadcast. At the bottleneck itself, no recommendation — vehicles can accelerate into the freed space.
Immediately upstream of the bottleneck, the recommended speed is v_up = β · v_m, where β < 1 controls aggressiveness. CVs receive RSMs at 0.1 Hz, hold the new limit for 10 s, then refresh or revert.
Shared with Queue Warning: 10 weaving sections, on/off ramp pair 1,000 ft apart, upstream and downstream VDS at every section. Ramp meters installed at every on-ramp.
On-ramp demand bumped to 900 vph at section #10 (vs. 300 vph elsewhere). Metering rate at section #10 set to 900 vph to allow that demand through. The forced bottleneck propagates upstream.
Across every β setting, demand level, and CV penetration rate tested, the implemented Speed Harmonization algorithm failed to improve mobility — and in many cases increased average delay.
β = 0.85, 0.90, and 0.95 all produced equal-or-worse system delay than the baseline. The aggressive reduction (β = 0.85) was the most damaging.
Extending RSU range from 1,640 ft to 2,460 ft (750 m) at β = 0.95 produced essentially identical results. The algorithm — not the communications — is the limit.
The core mechanism failed: slowing upstream traffic just before a weaving bottleneck did not create enough space for merging flows to exit faster. Queue Warning's results confirm the same structural issue.
Because the algorithm itself doesn't deliver benefit, increasing the share of compliant CVs from 10% to 50% just amplifies the delay penalty.
Speed Harmonization is not safety-critical and the algorithm aggregates data over much longer windows than communication delay. Communications quality was set aside as a non-factor.
Heavy weaving flows at on-ramps create the capacity drop. Reducing upstream throughput is the wrong tool — re-routing the on-ramp demand is the right one.
Speed Harmonization is a popular concept that doesn't survive contact with an actuated freeway weaving bottleneck. The mechanism is mismatched to the physics — and the solution lies elsewhere on the network.
The simulation results are blunt: every parameter setting tested left the system worse off than baseline. This isn't a failure of tuning. It's a structural mismatch — reducing inflow upstream does not create space for downstream weaving flows to clear, because the bottleneck isn't capacity-limited by upstream supply, it's limited by the merge geometry.
Queue Warning's results corroborate this. Slowing upstream vehicles only helps when demand slightly exceeds reduced capacity — a narrow band, and not the regime where most bottlenecks operate. Outside that band, the speed reduction just becomes added delay.
The path forward is demand reassignment, not speed manipulation. Rerouting on-ramp flow at the bottleneck to upstream or downstream on-ramps directly addresses the merging conflict. This is Integrated Corridor Management territory — the next study, not this one.
Do not deploy this Speed Harmonization implementation for arterial-weaving freeway bottlenecks. Across all parameter settings tested, it increased average vehicle delay.
For heavy weaving flows, evaluate ramp-flow rerouting instead — directing on-ramp demand to upstream or downstream on-ramps with available capacity.
If pursuing speed harmonization research, focus on bottleneck types other than weaving (incidents, work zones, lane drops) where reducing inflow may genuinely create downstream space.
Don't invest in extended communications range or higher CV penetration for this application. The bottleneck isn't communications-bound; the algorithm is.
Keep the V2X Microsimulation Platform's Speed Harmonization scaffolding intact. Future algorithm variants — including ML-driven or collaborative-control approaches — can be plugged in and re-tested on the same scenarios.