Software / What’s new in 11 / Network convergence

Convergence you can trust.

Stable results for oversaturated networks. The Enhanced Network Convergence Method in SIDRA INTERSECTION 11 adds a Two-Loop architecture alongside the existing One-Loop method, with diagnostics that show whether results have settled.

Inner Loop → Outer Loop → Stopping condition → ConvergedTwo-Loop method · Lane blockage held constant in the Inner Loop, recalculated in the Outer Loop
01 / Why convergence matters

Stability, reliability, auditability.

Unsettled results cannot be audited

When network model iterations fail to settle, results become unreliable, and unreliable results cannot be used for professional audits, infrastructure planning, or government submissions.

In any network model, the iterative process must resolve complex interdependencies between sites: lane-by-lane capacity with shared and opposed turns, signal timing, queue interactions, and downstream queue blockage effects on upstream lanes. Under congested and oversaturated conditions, these iterations can become unstable, producing results that oscillate from iteration to iteration without settling.

Three causes3
01Queue-capacity feedback loopsA downstream queue blocks an upstream lane, reducing its capacity. The reduced capacity changes the queue pattern, which resets the blockage, creating oscillations that traditional single-loop methods struggle to resolve.
02The lane blockage challengeLane blockage is a structural constraint: the physical reality of a queue from one intersection preventing flow at another. It behaves differently from performance variables like capacity and signal timing, and needs a more deliberate resolution process.
03The cost of non-convergenceWhen the stopping condition is not satisfied, the maximum number of iterations is reached and results are unsettled. Large values of the Network Model Variability Index indicate high variability during iterations, and that final results may not be reliable.
02 / Two-Loop method

A new architecture for difficult networks.

Inner Loop · Outer Loop · Stable equilibrium

Traditional single-loop models attempt to resolve all network variables simultaneously. SIDRA INTERSECTION 11 separates the problem, solving performance variables and structural constraints in dedicated loops for superior numerical stability. By decoupling lane blockage constraints from capacity and timing calculations, the Two-Loop method achieves convergence in cases where single-loop approaches produce unsettled results, particularly in networks with high levels of lane blockage causing oversaturation. The Final Run step used in earlier versions after network model iterations has been removed.

Inner LoopPerformance processorHolds Lane Blockage Adjustment Factors constant, as passed from the Outer Loop, and iterates to resolve all other network variables: lane-by-lane capacity with shared and opposed turns, signal timing, and queue dynamics. Settles into a local equilibrium without the noise of fluctuating blockage values.Default 10 iterations · range 5 to 60
Outer LoopStructural validatorReceives updated lane blockage data, degree of saturation and queue storage ratio values from the converged Inner Loop. Recalculates weighted Lane Blockage Adjustment Factors and checks overall convergence. If factors have changed significantly, a new Inner Loop cycle is initiated.Default 30 iterations · range 5 to 100
StopStopping conditionIterations stop when lane degree of saturation and queue storage ratio values do not change by more than the Adjusted Percentage Stopping Condition across three consecutive iterations, or when the maximum iteration count is reached. Adjusted values are more tolerant at lower degrees of saturation.Default 1% · range 0.1% to 10%
03 / Convergence diagnostics

Know when to trust your results.

Network Summary · Route Summary · Diagnostics

Most tools provide a binary converged or not converged status. SIDRA INTERSECTION 11 provides granular metrics so you can diagnose exactly what is happening during network model iterations.

01Network Model Variability IndexThe average value of the largest changes in lane degrees of saturation or queue storage ratios from the third to the last network iteration. High values indicate that results may be unsettled and should be interpreted with caution.Uncertainty indicator
02Iteration-by-iteration reportingThe largest change in lane degrees of saturation or queue storage ratios for the last three iterations, with the change in Network Control Delay (average) and Network Speed Efficiency, so you can see whether results are trending towards stability or oscillating.Last three iterations
03Convergence method identificationOutput reports identify which method was used (One-Loop or Two-Loop), the number of iterations completed versus the maximum specified, and the stopping condition applied.Audit trail
04Adjusted Percentage Stopping ConditionThe program calculates adjusted stopping condition values based on the degree of saturation and queue storage ratio, applying more tolerant thresholds at lower saturation levels. The relationship is documented.Documented
04 / Analytical precision

Deterministic results, no seed runs.

Repeatable, auditable, one run
MicrosimulationSIDRA INTERSECTION 11
Runs neededMultiple random seed runs; the number needed grows with congestion and randomness.One run, with the same result every time.
Same inputsTwo engineers with different seed counts or warm-up periods may reach different conclusions.The same output every time.
Lane detailVehicle trajectories with random variation.Capacity, performance and blockage resolved at the individual lane level: short turn bays, shared lanes, merge points.
TraceabilitySimulation black box.Every parameter, stopping condition and metric is documented and reported. Defensible in audits.
05 / Traffic assignment

Route choice meets lane-level performance.

Assignment and network convergence in one platform

Most workflows require separate tools for traffic assignment and intersection analysis. SIDRA INTERSECTION 11 integrates both in a single environment where route choices are informed by detailed lane-by-lane performance.

SIDRA ASSIGN uses Origin-Destination trip demand data to estimate intersection movement volumes while accounting for network-wide traffic interactions and detailed intersection capacity and performance estimation at the lane level. It has its own convergence metrics.

Assignment convergence3
01Five assignment methodsStochastic User Equilibrium (default), User Equilibrium, Fixed Route Assignment, Incremental Assignment and All-or-Nothing, each with its own parameters. Duality Gap is the stopping condition for User Equilibrium and Stochastic User Equilibrium.
02Duality Gap convergenceA proximity metric measuring how close the current assignment is to a true equilibrium where no user can improve their travel time by unilaterally changing routes. Default tolerance 1%.
03Stability convergence metricsConsistency Index, absolute total travel time change, maximum change in route flows, and maximum change in site movement flows: confirmation that results are not just mathematically accurate but operationally stable.
06 / Practical guidance

One-Loop for standard cases, Two-Loop for difficult ones.

Network Data → Network Analysis Settings
MethodUse whenIterationsStopping condition
01One-LoopThe default. Resolves all network variable interactions simultaneously in a single iterative loop. Suitable for the majority of network analyses.Maximum 30
Range 5 to 100
1%
Range 0.1% to 10%
02Two-LoopUnsettled results with high Network Model Variability Index values after the maximum iterations, even at the largest setting. May take longer to process in some cases.Inner 10, Outer 30
Ranges 5 to 60, 5 to 100
1%
Adjusted by saturation

Balancing precision and processing time: when using demand and sensitivity analysis or optimum cycle time options for network signal timing calculations, the maximum number of iterations can be decreased and the stopping condition increased to reduce processing time. Reinstate the defaults once the network configuration stage is completed.

SIDRA INTERSECTION 11

Convergence is one part of the network story.

Discover how SIDRA ASSIGN brings micro-analytical traffic assignment to the Network model
Included in every subscription · Five methods