Under the night sky, the dock is brightly lit in the central control room. The dispatchers are arranging the work schedule for the next day or even several days: they need to find the optimal solution between the ship berthing window, quay crane resource allocation, yard container space planning, and loading plan. Each link is interrelated and interconnected, with each link causing a ripple effect. Improper arrangement in any area can amplify the chain reaction and lead to fluctuations in operational efficiency throughout Hong Kong. For a long time,The core ability of "scheduling" highly relies on the personal experience of senior dispatchers. Developing a high-quality global plan often takes several hours and is difficult to effectively reuse among different personnel and terminals. This is not only a pain point for individual ports, but also a common challenge faced by the global port industry.

At present, the automation of dock equipment has been widely popularized, but the production command is limitedThe "brain" still relies mainly on manual experience judgment: key businesses such as berth allocation, ship piloting, yard layout, and container loading have not yet been freed from manual dominance, and various business systems are fragmented, forming "data islands" that are difficult to fully unleash the value of data.The next stage of port intelligence has shifted its focus from incremental investment in hardware equipment to building smarter "decision brains" - transforming the experience of senior experts into reusable algorithmic intelligence, and transforming dispersed data into a global optimization perspective。

The key to solving the problem,To build a set that meets bothThe evolution law of AI technology, combined with the engineering system that fits the characteristics of port business, enables AI to truly "understand ports and be able to land"Proposed at the 2026 Huawei Connect ConferenceDIMAK Engineering SystemIt is born for this purpose and includes five core competency dimensions:

• Data Engineering（Data Engineering）Integrating heterogeneous data from multiple sources such as devices, business, and video perception, governance becomes a high-quality enterprise data asset available for AI.

   

• Infrastructure Engineering（Infrastructure Engineering）Provide an open, compatible, and scalable computing power base, offering cloud edge collaboration, resource scheduling, security isolation, and other capabilities to support high concurrency requirements for large model training and inference.

   

• Model Engineering（Model Engineering）By means of fine-tuning, distillation, quantification, and full chain evaluation, the basic large-scale model is transformed into a model asset for the port industry, significantly improving the performance of general models in understanding port professional terminology and complex business logic reasoning.

   

• Intelligent Agent Engineering（Agent Engineering）Design and govern intelligent agents with autonomous action capabilities around enterprise business processes, clarify permissions, responsibilities, and human-machine collaboration boundaries.

   

• Knowledge Engineering（Knowledge Engineering）Build a business semantic foundation for mutual understanding between humans and machines by standardizing enterprise systems, SOP standards, and expert implicit experience.

DIMAKThe core value of the system lies in enablingAI is useful and applicable in the industryfast、Utilize value。It is not intended to replace the original dockTOS, ECS, and other core systems serve as the foundation for AI capabilities and coexist with them, allowing existing docks to be built without having to be overturned,Can achieve intelligent upgrade at a lower cost。

Image:Multi agent collaboration in portsDIMAKArchitecture

Meanwhile, based onThe DIMAK system, jointly created by Huawei and port customers and partners, has accumulated a set of practical and replicable port AI implementation paths. Taking container terminals as an example, although there are differences in hardware configuration and management details among ports, their core business processes and decision-making logic have a high degree of industry commonality.Through this "commonality abstraction+individuality adaptation" model, Huawei's solution validated at a benchmark port can achieve seamless reuse across ports with over 80% of its core capabilities, requiring only lightweight customization for the remaining local featuresThis not only significantly reduces the trial and error costs and construction cycle of a single terminal, but also breaks the industry's dilemma of "repeated wheel making". Huawei is committed to opening up and contributing to this validated implementation experience, toolchain, and process framework, enabling global port operators to quickly build intelligent solutions adapted to local characteristics based on a unified base, thereby forming an industrial synergy and accelerating the large-scale landing of AI technology in the port industry.

The DIMAK system, as a universal AI engineering foundation, maintains loose coupling with specific business scenarios and has strong adaptability to different scenarios.Huawei has chosen the typical scenario of "port production planning" as a breakthrough point, and based on this system, has created a plug-in layered intelligent agent architecture, achieving end-to-end implementation of the "5+1" plan.The five specialized intelligent agents of pilotage, berth, quay crane, yard, and loading each perform their respective duties, undertaking tasks such as ship pilotage scheduling, berth allocation, quay crane operation layout, yard container space planning, and container loading; As the "smart brain" that coordinates the five major plans, the all factor scheduling intelligent agent integrates variables such as arrival fluctuations, equipment failures, and tidal constraints to perform conflict verification and global optimization on various specialized plans, covering the entire production planning process. For the dock, the planning position has bid farewell to "fighting on its own" and moved towards "playing a game of chess as a whole" - adding a tireless and comprehensive "commander-in-chief".

The value created by this model for the port is becoming apparent. The berth plan has been compressed from hourly level to minute level, and the acceptance rate of the operation plan has exceeded90%, the optimal plan generated by AI is no longer just for reference only,But it can directly issue precise instructions to the shore bridge, yard bridge, and horizontal transportation equipment for execution, formingThe complete closed loop of "perception inference decision execution",Transforming intelligent investment directly into increased throughput and reduced operating costs。

Of course, port intelligence is not something that can be achieved overnightEspecially in the renovation of existing docks, there are challenges such as the coexistence of old and new business systems, difficulty in data governance, highly personalized scenarios, and scarcity of composite talents. In practice, we have found that some projects are affected byFocusing on applications and neglecting the accumulation of computing power, data, and industry knowledge has led to difficulties in iterative expansion after launch. Based on practice, we propose an overall implementation strategy for the intelligent transformation of ports to ensure a stable and far-reaching path of transformation

• Base first, layered construction: PersistPrioritize 'foundation', first solidify the computing power infrastructure, complete data governance, accumulate industry knowledge base, and then gradually stack intelligent applications. Avoid castle in the air development and ensure that the system has long-term evolution capabilities.

   

• Value oriented, take small steps and run fastChoose the most painful and effective scenarios (such as planning and intelligent security) to enter, verify the effectiveness through small-scale pilot projects, build confidence, and then evolve towards full element collaboration.

   

• Open collaboration, ecological co constructionBreaking the blockade, allowing port operators, equipment manufacturers, research institutions, andICT partners participate together. Only when advanced practices flow in the industry chain can the overall level of the industry be accelerated.

   

Artificial intelligence is profoundly reshaping the production paradigm of ports, creating tangible value for the global port industry——Faster plan response, more stable work order, lower operating costs, and safer on-site environment. As of now, Huawei has served over 100 ports worldwide. Looking forward to the future, we look forward to working together with global port operators and industry chain partners to enable more ports to have increasingly accurate "AI brains", promote ports from single point intelligence to full chain intelligent collaboration, contribute intelligence and strength to the digital transformation and green sustainable development of global ports, and provide Chinese solutions for the construction of world-class ports.

END

 

 

Previous hot articles