CITSE Exhibitor Recommendation|
SUTPC Presents ShenYan Transportation Large Model and World Premiere of L4 CityLoog Autonomous Minibus at CITSE 2026!
CITSE 2026 revolves around the core theme of "Focus on Industrial Innovation and Transformation, Empower the Construction of Overseas Expansion Ecology" and will grandly open at the Xiamen International Conference & Exhibition Center from May 13 to 15.
Gathering cutting-edge insights and leading transportation transformation, the New Product Launch, as the core engine event of the CITSE 2026, will provide global enterprises with an authoritative platform for premieres and first shows. It will centrally showcase cutting-edge innovative technological achievements and solutions in the field of smart transportation, including AI+, digital highway collaboration, autonomous driving, traffic signal control systems, urban smart rail transit, and transportation-energy integration. Participants can engage in face-to-face exchanges with potential clients and partners from around the world to explore cooperation opportunities.
CITSE 2026 features a major exhibitor—Shenzhen Urban Transport Planning Center Co., Ltd. (hereinafter referred to as "SUTPC"), which will present a series of cutting-edge achievements, including the globally launched ShenYan Transportation Large Mode, the L4 CityLoog autonomous minibus, an embodied intelligent unmanned delivery vehicle, and the TransPaaS platform.
SUTPC was founded in 1996 and listed on the ChiNext of the Shenzhen Stock Exchange in 2021 (Stock Code: 301091.SZ). It is a world-leading provider of new-generation integrated urban transportation solutions and intelligent products.
By focusing on AI-driven development, SUTPC promotes the continuous evolution of the TransPaaS platform, supports the construction of an integrated air-ground intelligent transportation system, and improves the overall efficiency of urban transportation. As a national high-tech enterprise and a National Enterprise Technology Center, SUTPC has assembled a multidisciplinary team of nearly 2,000 professionals. Its business covers more than 160 key cities across 30 provinces and municipalities in China, as well as over 10 international cities including Hong Kong, Singapore and Abu Dhabi. Its controlling shareholder is Shenzhen Smart City Technology Development Group Co., Ltd., which, as the chain leader of the industrial innovation chain, coordinates the construction and operation of smart city and digital government initiatives in Shenzhen.
Three Major New Products Make a Powerful Debut
1. ShenYan Transportation Large Model: A trillion-parameter industry-specific large language model launched globally
At CITSE 2026, SUTPC officially launches the ShenYan Transportation Large Model for global users and developers, initially opening up five cities for user experience. This model is an industry foundation model for comprehensive urban transportation, featuring a trillion-parameter scale. It possesses four core capabilities: graph-enhanced cross-domain fusion reasoning, anomaly-state precise prediction through large-and-small model integration, macro-micro multi-scale computation driven by multiple agents, and adaptive dynamic control driven by reinforcement learning.
Core Highlights:
- The industry foundation model that best understands transportation: Built on 770 billion traffic data points covering nearly 300 cities across the country, it features hundreds of billions of parameters and can deeply understand the complex spatiotemporal relationships of urban traffic.
- Four core capabilities: Possesses cross-domain fusion reasoning, anomaly-state precise prediction, macro-micro multi-scale computation, and adaptive dynamic control capabilities, effectively addressing industry challenges such as cross-domain integration difficulties, multi-agent collaboration difficulties, and anomaly intervention difficulties.
- Verified practical effectiveness: In Shenzhen's Futian Central Area, peak-hour vehicle speeds increased by approximately 7.5%; in Guangdong-Hong Kong-Macao Greater Bay Area highway prediction scenarios, prediction accuracy reached 93.7%, a 25%+ improvement over traditional methods; emergency response plan generation was compressed from hours to minutes.
- Global application practice: Has been piloted in over 20 domestic cities and international markets such as Abu Dhabi and Singapore, with significant results.
Future Plan:
SUTPC will continue to deepen the integration of the ShenYan Transportation Large Model with data on urban development, industrial economy, meteorology, environment, etc., upgrading to the next-generation smart transportation operating system TransPaaS platform centered on the traffic foundation model, agents, and embodied intelligent equipment. It will expand multi-modal collaborative applications in low-altitude economy and autonomous driving, promoting the shift of urban traffic governance from "passive response" to "active governance."
2. L4 CityLoog Autonomous Minibus
As the nation’s first mass-produced Level 4 autonomous minibus designed for open-road operation and capable of large-scale commercial deployment. Leveraging the autonomous bus operation and testing infrastructure in Hong Kong and Shenzhen, and utilizing a technology demonstration route at a Shenzhen park as its platform, the vehicle will serve as an international showcase for major APEC diplomatic events. Its launch marks China’s autonomous minibuses officially entering a new phase of “driverless, routine, and commercial” operations, establishing replicable and scalable technical solutions and operational standards that provide critical support for the leading development of smart transportation and the intelligent connected vehicle industry in the Greater Bay Area.
The L4 CityLoog autonomous minibus is planned to begin trial operation on a Shenzhen park ring road in 2026. It adopts domestic high-compute chips and a fully redundant safety architecture, with a standard remote driving system (video streaming latency<200ms), supporting a "demand-responsive, precisely scheduled" all-weather operation model. The project aims to achieve fully driverless commercial operation within three years, scaling up applications in scenarios such as urban shuttle, hub distribution, campus micro-circulation, and low-altitude economy support, contributing to the low-carbon and intelligent upgrade of urban transportation.
3. Embodied Intelligent Unmanned Delivery Vehicle
Targeting last-mile logistics scenarios in urban buildings, Shenzhen Urban Transportation has launched an embodied intelligent unmanned delivery vehicle that integrates an autonomous driving chassis, an embodied intelligent robotic arm, and a cloud-based collaborative dispatch platform to enable access through security gates, movement between floors, autonomous loading and unloading, and precise interaction.
Core Highlights:
- Autonomous driving system is based on ROS2/Apollo and supports autonomous navigation across access control points and floors;
- Embodied intelligence operation layer: Equipped with a robotic arm, supporting autonomous loading/unloading, opening doors, calling elevators, pressing floor buttons, and other physical interactions.
- Cloud collaborative dispatch layer: Supports order management, task scheduling, remote monitoring, and system upgrades, ensuring scalable implementation. This product embodies the technological evolution of autonomous delivery from merely “being able to move” to “being able to operate, deliver, and implement.”
Application Scenarios:
- Last-mile delivery in industrial parks/office buildings
- Indoor delivery in hotels and commercial office buildings
- Unmanned delivery in semi-enclosed scenarios such as hospitals, exhibition halls, and campuses
Core Exhibit: TransPaaS Platform
The TransPaaS platform is a next-generation smart transportation operating system based on new technologies such as foundation models, embodied intelligence, and digital twins. Through large-and-small model nesting and multi-level integrated control, it achieves global traffic network balance and local autonomous optimization.
Key Technologies:
✔ LLM Training & Inference with Reasoning Enhancement & AgentRL
ReFT+GRPO powers a full-stack traffic-domain training system for domain knowledge alignment. AgentRL + multimodal attention enables hierarchical agents for complex decision-making. INT4/AWQ + Sparse Attention ensures low-latency inference, closing the train-infer-apply loop.
✔ MCP+A2A Cross-Domain Agent Collaboration
Based on Linux Foundation’s MCP standard, unifies 5 connection types, integrates 17K+ open-source MCP servers and 350+ enterprise data sources. A2A bidirectional streaming enables cross-domain multi-agent networks.
✔ Hierarchical RAG & Dynamic Agent Skill Orchestration
Four-tier knowledge base (regulations, standards, cases, terminology: 360+ items) supports RAG evolution (vector search → KG → fine-tuning). SkillOrchestrator engine with DSL enables dynamic orchestration of 63 skills + 41 tools.
✔ Efficient Fusion of Large & Small Models
Large models handle global reasoning, small models focus on edge/lightweight tasks. Dynamic routing, hierarchical inference, cascade generation, knowledge distillation, and cloud-edge synergy deliver low latency, high throughput, and cost-effective AI for millions of agents and millisecond-level device control.
✔ Full-Element Digital Twin & Efficient Rendering
Covers CIM, BIM, IoT. Manages tens of millions of lightweight models with millisecond high-precision rendering, supporting millions of CIM/BIM model loads.
✔ HarmonyOS-Based Heterogeneous Multi-Protocol Embodied AI Swarm Control
Based on HarmonyOS UDH unified interconnection framework, adapts 10+ device protocols (NTCIP/V2X/MQTT/ONVIF/CV2X/ROS2/MAVLink), enabling unified access and coordinated control of embodied agents across vendors, protocols, and scenarios.
Core Capabilities:
1. Agent Models:
Based on a scientific computation foundation model for the transportation industry, guided by business applications, and combined with the translation of scientific research results, SUTPC has constructed an agent system featuring "triadic collaboration" of Behavioral Agents, Operational Agents, and Control Agents.
(1) Behavioral Agents: Using AI and digital twins, digitally model the identity and behavior of pedestrians, vehicles, roads, environments, and facilities to accurately replicate their physical actions and mechanisms in a digital twin world.
(2) Operational Agents: Using MLP for deep perception of physical entities and real-time business status, train expert sub-models, and leverage MoE dynamic routing to efficiently handle complex tasks, enabling accurate state recognition and problem diagnosis.
(3) Control Agents: Through inference optimization fine-tuning, using INT4/AWQ quantization, structured pruning, KV Cache management, and TensorRT-LLM compilation acceleration, they achieve end-to-end inference latency stably ≤20 seconds under long-context input scenarios.
2. Embodied Robots:
Leveraging the distributed collaboration capabilities of the HarmonyOS operating system, SUTPC integrates over 220 embodied robots including drones, robotic dogs, unmanned vehicles, and intelligent inspection vehicles. It deeply integrates VSLAM visual positioning and SLAM simultaneous mapping and localization technologies, relying on world foundation model technology for global perception and intelligent decision-making, to achieve autonomous path planning, autonomous precise navigation, and dynamic coordinated control of embodied robots.
3. Platform Tooling:
Aiming at new cities, new projects, and new scenarios, SUTPC has developed an agile and efficient toolset to support the large-scale promotion of the platform. It supports one-click generation of city twin maps, 3D reconstruction, digital twin scenes, and business application orchestration.
4. Data & Corpus:
SUTPC has built a professional corpus and knowledge base for the transportation domain, including standards and specifications (over 2,500 national, industry, local, etc. standards), manuals and textbooks (over 500 items including the Highway Capacity Manual (HCM), Road Entrance/Exit Management Manual, text reports, etc.), totaling 700 GB and 30.6 billion entries.
To experience the unique charm of SUTPC's cutting-edge smart transportation technologies and products up close and to discuss cooperation opportunities face-to-face with the SUTPC technical team, please visit the Xiamen International Conference & Exhibition Center from May 13 to 15 and explore the Future Transportation Exhibition Area of the CITSE 2026. Here, you can not only witness the spectacular debut of globally premiered innovative technologies but also deeply engage with industry-leading enterprises to jointly explore the future development path of smart transportation.
