[MOVING TOWARDS GREATER SAFETY]
Conflict Analysis Engine
QUANTIFY RISKS // PREVENT CRASHES // OPTIMIZE INFRASTRUCTURE
Roadscor’s Conflict Analysis Engine operationalizes forward‑looking, simulation‑based and conflict‑based methodology to quantify road‑user risk in targeted road layouts (e.g., intersections). Evolving from methodology originally developed by Siemens, the engine discovers known and unknown unsafe scenarios by simulating all feasible movements and systematically exploring plausible but previously unseen conflicts.
Step 1 — Data Collection
We ingest time‑aligned trajectories of all interacting road users together with traffic‑light/priority states to reconstruct temporally consistent interactions in simulation.
- Trajectories: Road‑user trajectories for all interacting actors in the region of interest, so that rare combinations and corner‑case behaviors can be reconstructed and replayed in simulation.
- Priority Scheduling: Traffic‑light/priority status over time, enabling time‑synchronized reconstructions where hazards emerge from timing and compliance variations.
Capturing the right temporal context enables us to find these cases systematically rather than retrospectively.
Step 2 — Feature Extraction (Graph Model & Actor Distributions)
We transform each selected road layout into a graph ( edges = feasible maneuver segments, nodes = start and end of the maneuver segments) to structurally encode where and how actors move and interact. From the collected trajectories, we compute actor‑level behavioral features—speed profiles, lateral offsets, timing/phase behaviors, and compliance tendencies—and aggregate them into empirical probability distributions tied to the maneuvers. These distributions quantify what typically happens and constrain later search to realistic behaviors.
Step 3 — Optimization (Discovering Known & Unknown Unsafe Scenarios)
We cast scenario discovery as a constrained optimization over a decision vector (paths, timings, speeds, lateral offsets, actor presence/roles), rather than using inefficient randomization to find conflicting scenarios. The optimizer searches for boundary conditions of safe operation—where near‑misses and violations appear—while remaining within the realistic envelope defined by observed behavior.
Objective & ranking – Candidate scenarios are evaluated using a cost function that reflects Roadscor’s conflict‑based philosophy: Criticality of the interaction, Probability derived from actor/scenario likelihoods, and Controllability reflecting maneuver margins to avoid collision.
The optimizer emphasizes scenarios that increase criticality and decreases controllability without collapsing probability, producing a ranked set of known and unknown unsafe cases for scoring the safety of the road layout.
Beyond Static Checks
Because users are evaluated against each other in time, Roadscor exposes ‘invisible’ risks that purely geometric checks miss – such as phase offsets, platoon dispersion, and complex yield negotiations. The result is a forward-looking risk profile delivered with probabilistic confidence and a clear audit trail.