Make a step to KKW implementing VDR

[LdDl]

WORK IN PROGRES

Removes the unfinished, mis-calibrated KKW (Kerner–Klenov–Wolf ) layer and replaces it with VDR (Velocity-Dependent Randomization) baseline plus the full driver-behaviour machinery (aggression gradient, patience, lane-change safety) that a future KKW layer will build on.

1. VDR

The longitudinal model is now plain NaSch + two-level VDR: a moving-dawdle probability p and a higher slow-to-start probability p0 while stopped (p0 ≥ p). This produces the realistic capacity drop / metastability - the foundation KKW is built on.

2. Driver style as one continuous gradient

Aggressiveness is now a continuous aggressive_level = 1 - cooperativity (no longer an enum type). Behaviour presets derive change_p1 (the probability of actually committing an allowed lane change) and lc_cooldown from it.

3. Cut-in aggressors

New gradient aggressor_advantage (constant threshold 0.8): an aggressive driver cuts in - bypasses the rear-safety gap at intention time and wins a contested cell against any less-aggressive vehicle, forcing it to brake. Applied uniformly across all non-physical conflict rules (merge-lane-change, merge-forward, forward-lane-change, cross-lane-change). Replaces the old coarse "Aggressive-enum vs Cooperative-enum" checks.

4. Patience / desperate override (deadlock escape)

A vehicle starved past patience() (300–700 steps, scaled by cooperativity) forces its way through - it wins any non-physical contested cell. The most-starved vehicle wins, ties broken by lowest id; it never pushes through a physical tail (a body); the one-occupant-per-cell invariant is preserved (claimed-cells + occupancy check + sweep-through clamp).

5. Lane-change safety & anti-weaving

Speed-aware rear gap (rear_safe, using a start-of-step speed snapshot) + front gap (clear_forward) + an incentive gate + a stochastic P1 commit. Plus a per-type reactive lane-change cooldown (timer_reactive_lane_change / lc_cooldown).

6. Determinism

Seedable RNG (MTSC_SEED); all stochastic draws are unconditional and in a fixed per-vehicle order. A* tie-break via total_cmp (NaN-safe). The full-tie conflict tie-break went from a coin flip to lowest vehicle id (invariant to participant order). Result: under a fixed seed its runs in a reproducible way.

8. Tests

Extensive new coverage: desperate override (across every conflict type), cut-in aggressor (forward and crossing), order-invariance of the lowest-id tie-break, the P1-vs-aggressiveness mapping, and determinism.

References

Cellular-automata traffic model (Keldysh Institute of Applied Mathematics, RAS):

• A. A. Chechina, N. G. Churbanova, M. A. Trapeznikova. A cellular automata model for describing mixed flow of passenger cars and trucks on multilane highways // Computer Research and Modeling, 2026, Vol. 18, No. 1, pp. 61–80. https://doi.org/10.20537/2076-7633-2026-18-1-61-80
• A. A. Chechina. Driver behaviour algorithms for overcoming unsignalized intersections with priority pass and for bypassing an obstacle // Matematicheskoe Modelirovanie, 2021, Vol. 33, No. 9, pp. 47–59. https://doi.org/10.20948/mm-2021-09-04
• A. A. Chechina, N. G. Churbanova, M. A. Trapeznikova. Comparison of the reproduction of spatiotemporal traffic-flow structures using different data-averaging methods // Matematicheskoe Modelirovanie, 2021, Vol. 33, No. 1, pp. 25–35. https://doi.org/10.20948/mm-2021-01-02
• N. G. Churbanova, A. A. Chechina, M. A. Trapeznikova, P. A. Sokolov. Simulation of traffic flows on road segments using cellular automata theory and quasigasdynamic approach // Mathematica Montisnigri, 2019, Vol. XLVI, pp. 72–90. https://doi.org/10.20948/mathmontis-2019-46-7
• M. A. Trapeznikova, I. R. Furmanov, N. G. Churbanova, R. Lipp. Simulation of multilane vehicle traffic flows based on the cellular automata theory // Matematicheskoe Modelirovanie, 2011, Vol. 23, No. 6, pp. 133–146.
• M. A. Trapeznikova, A. A. Chechina, N. G. Churbanova. Two-dimensional cellular automata model for simulation of traffic flows on the elements of a city road network // Matematicheskoe Modelirovanie, 2017, Vol. 29, No. 9, pp. 110–120.