1. Challenges

Traditional intersection management suffers from fragmented perception and rigid control. Existing facilities—signal controllers, enforcement cameras, and surveillance—often operate in isolation, producing sectional and fragmented data that cannot support complete vehicle trajectory reconstruction. Signal control largely relies on fixed timing plans based on historical statistics, lacking the ability to respond dynamically to real‑time traffic variations. Moreover, management decisions on congestion causes and safety hazards are constrained by limited data dimensions, making accurate diagnosis and science‑based decision‑making difficult. Urban traffic management urgently needs a systematic shift from experience‑driven to data‑driven approaches.

2. Our Solution

This solution builds a holographic sensing system centered on wide‑area radar, upgrading the front‑end collection model to provide precise, real‑time, and full‑element perception data for the signal control system. Leveraging diversified radar data, it establishes a new detection model that forms the data foundation for signal timing optimization. A signal evaluation system is also implemented to quantify operational performance, creating a closed‑loop mechanism of "sensing – evaluation – optimization". This provides accurate data support for signal control, traffic guidance, and command & dispatch platforms, enabling regional adaptive control and improving both the intelligence of signal control and the effectiveness of global data utilization.

2.1 System Composition

The system consists of three layers: front‑end sensing, data processing, and central application.

• Front‑end sensing layer: Millimeter‑wave radar collects real‑time traffic data in all directions, including vehicle position, speed, headway, queue length, and lane occupancy—forming a high‑precision, all‑weather perception data source.

• Data processing layer: Multi‑dimensional radar data is cleaned, fused, and structured to extract key indicators such as lane‑level queue length, per‑lane traffic volume, and area status. Standard interfaces (RS485, I/O, RJ45) enable data exchange with the signal control system and big data platform.

• Central application layer: The big data platform aggregates radar, signal, and other multi‑source data, performs spatial‑temporal alignment and data governance, and builds a signal control rating system that reflects real‑time network status and traffic demand changes—providing data support for the signal optimization team.

2.2 How it works

The system uses millimeter‑wave radar as the core sensing unit to capture real‑time vehicle position, speed, queue length, lane occupancy, and other microscopic parameters in all directions—delivering a high‑precision, all‑weather perception data source. The radar scans at 20 Hz, continuously outputting lane‑level queue length, vehicle count, target trajectory, and area status through multi‑section detection zones across multiple lanes. After structuring, the data is pushed via standard interfaces to the signal control system and big data platform in real time, supporting signal timing optimization, operational evaluation, and control strategy adjustment—forming a full‑loop closed chain of "sensing – evaluation – optimization".

2.3 System Functions

• Large‑area holographic sensing for single‑point adaptive control – Based on a fine‑grained multi‑section, multi‑lane detection architecture, the radar provides wide‑area, high‑precision coverage of all intersection approaches, outputting lane‑level queue length, vehicle count, target trajectory, and area status—delivering accurate perception input for intelligent phase early termination and extension.

• Multi‑section detection data for arterial coordination – With multi‑section detection across a large area, the radar outputs per‑lane traffic volume, queue length, travel time, and other accurate parameters—supporting cycle prediction, split adjustment, and offset optimization for dynamic green wave coordination on arterials.

• High‑precision presence and queue detection for area‑wide coordination – The radar provides lane‑level presence and queue status, outputting queue length, queuing vehicle count, and head/tail positions per direction. The signal system uses this data to grasp real‑time traffic demand and achieve coordinated control from single intersections to the entire network. Queue detection also enables advanced applications such as spillback prevention, ensuring timely signal response when queues approach overflow.

• Operational evaluation for signal strategy iteration – Based on radar‑collected indicators such as number of stops, delay, and queue length, the system quantitatively evaluates timing performance—providing data‑driven evidence for the optimization team to support continuous iteration and fine‑tuning of signal control strategies.

3. Core Values

• Diversified perception data – With deep radar‑video fusion as the core, the system enables large‑area, all‑weather, holographic intersection sensing. Fine‑grained multi‑section, multi‑lane detection delivers lane‑level queue, volume, trajectory, and area status—providing comprehensive, blind‑spot‑free multi‑dimensional data support.

• High‑reliability detection accuracy – Unaffected by lighting, rain, snow, or fog, the system ensures 7×24 high‑precision presence and queue detection, offering stable and reliable perception data for signal control.

• Closed‑loop evaluation for continuous optimization – A full‑dimension signal evaluation system quantifies timing performance using objective data (stops, delay, etc.), providing accurate evidence for signal iteration—shifting signal timing from experience‑based adjustment to data‑driven fine‑tuning.