Transport for NSW engaged members of SoSavvy to craft a first port of all real-time traffic management response system in Australia, leveraging legacy roadway sensors and signalling protocols to produce intelligent incident response plans for incidients in major NSW roadways in real-time.
To enable real-time adaptive traffic management across NSW's arterial network, SoSavvy engineers contracted to Cubic to assist in integrating a modular Transport Management Platform (TMP) architecture—drawing inspiration from scalable systems like Cubic's TMP—for high-throughput, low-latency operations. By federating roadside sensor feeds, historical models, and live traffic telemetry into unified microservices, we established an orchestration layer that can handle thousands of events per second without bottlenecks.
Scalable Incident Detection & Prediction
The TMP system built a hybrid event pipeline: edge processors at signal boxes aggregated raw loop and video-sensor data, while upstream cloud modules ingested anomaly scores and queued event candidates. We layered a real-time incident detector powered by streaming analytics and trained models that learned typical flow patterns per corridor. Alerts propagate in under 250 ms, enabling the TMC to dispatch responses within seconds.
Dynamic Response & Route Guidance
Once an incident is flagged (slowing speeds, abrupt deceleration, occupancy surges), the platform automatically cross-references detour recommendations, variable message sign networks, and signal re-phasing commands. The system can issue route diversions or ramp-metering adjustments in real time, seamlessly scaling across multiple precincts. In high-priority corridors, this orchestration shortened simulated average incident clearance times by ~25%.
Continuous Learning & Feedback Loop
The TMP continuously ingests post-event telemetry (resolution times, flow-recovery gradients, driver behavior) to retrain detection thresholds and calibrate control heuristics. Each corridor becomes “smarter”—learning baseline patterns (weekday, peak vs off-peak, weather)—enabling more precise anomaly detection and fewer false positives. Over months, sensitivity improved while operator load decreased.
Key Capabilities
- Real-time camera integration: Native support for RTSP/ONVIF streams, edge motion/occlusion detection, frame hashing for event correlation, and privacy-aware snapshotting for operator consoles.
- Traffic control systems interoperability: Southbound adapters for established URBAN/UTC/SCATS-style interfaces (signal state, plans, phases) with northbound APIs to issue timing plan updates and VMS/ramp-meter commands.
- Legacy & emerging standards: Gateways for NTCIP/DATA-EX II/GTFS-RT feeds, archival loop/ANPR sensors, and modern CV/AI detectors—normalised into a common event schema for analytics and replay.
- Security by design: Zero-trust segmentation, mTLS between services, signed command payloads, per-corridor RBAC, audit trails, and isolated operator roles; secrets managed via HSM-backed KMS.
- Scale & resilience: Event-driven microservices with back-pressure control, horizontal autoscaling, multi-AZ HA, and idempotent command execution ensuring consistent outcomes during bursty incident waves.