How G7 Combines AI, Big Data, and IoT to Transform Logistics

Recently, the rise of 5G and increasing AI data demands have intensified challenges for engineers, prompting the question of how big data should be utilized and integrated with AI‑driven industries.

In response, Wang Gang, chief architect of the G7 foundational R&D platform, delivered a talk titled “G7 – AI + Big Data + IoT Architecture Practice” at the iResearch “Open Source & Ecosystem – Developer Session”.

What Is IoT Logistics?

IoT logistics refers to the real‑time management, tracking, and monitoring of transportation processes through Internet‑of‑Things technology.

G7 Platform Overview

G7 is a neutral, open technology platform that builds an industry‑wide ecosystem via an “AI + IA” (Artificial Intelligence + Intelligent Assets) strategy, serving over 60,000 customers and connecting more than 1.2 million vehicles.

The platform’s technical stack consists of a device layer, a multi‑platform gateway, a business platform, business systems, front‑end and mobile applications, and an independent operations system. Removing the device layer reveals an architecture similar to that of typical internet companies.

Big‑Data Infrastructure

At the bottom lies a standard open‑source big‑data stack, including an algorithm platform, business modeling, and a data warehouse. The algorithm platform sits beneath the model and warehouse to provide reusable core capabilities.

Data from vehicle devices are reported as messages following G7’s protocol. After gateway parsing, the messages are split into two pipelines and stored in HBase or Hive, following a Lambda‑style architecture.

Streaming with Kafka

Parsed messages are pushed to Kafka, which handles up to 900 k messages per second on average and peaks at 1.5 M messages per second. G7 operates a GMQ management system to ensure Kafka cluster stability and efficiency.

Real‑Time Processing with Flink

G7 uses Flink for stream processing, handling over 40 billion records (≈4 TB) daily. Events such as region entry/exit, fatigue detection, overspeed, and idling are computed in Flink, with a Glink management system overseeing task stability, resource isolation, and scheduling. Real‑time machine‑learning risk‑prediction algorithms are slated for future deployment.

Storage in HBase

Data stored in HBase follows five dimensions: who (device IMEI), when (timestamp), where (GPS), what (event), and how (processing details). End‑to‑end latency from reporting to storage is about 500 ms.

Visualization and Offline Analytics

A real‑time dashboard displays high‑risk vehicles and incident statistics, such as collisions and rollovers. Offline, G7 has implemented a patented unloading‑point recognition algorithm that virtually reconstructs cargo‑truck camera views to identify loading and unloading activities.

Safety Insurance Service

G7’s safety‑insurance business is device‑centric, combining devices, platform, humans, and AI services. In signal‑blind zones, the vehicle’s onboard system runs algorithms; AI intervenes first when an incident occurs, and human operators step in if AI fails.

Hardware Smart‑Wave Impact

Smart home, wearables, and other connected devices generate massive data, enabling insights into usage timing, feature adoption, and business value, which drive product improvement and growth.

Case Studies

Onecup – a smart kitchen appliance brand leveraging iResearch’s data platform for digital operations.

Anshida – a home‑appliance service provider using data intelligence to refine user consumption analysis and enhance on‑demand services.

For the full presentation slides, see the download links provided in the original source.