WorldCargo News

IGO Solutions and Avlino Inc have successfully integrated a range of AI modules for MICT in the Philippines

AI progress for container terminals

AI presents an interesting challenge for container terminal operators. The industry is still progressing on its journey with digitalisation, automation and optimisation. At the same time, AI offers the promise of making terminal operations much more efficient, both in terms of productivity and lowering costs. But how do container terminals even start with AI?

Working together, two sister companies are offering a way forward with an AI solution that “works as a TOS plugand-play accelerator”.

ICTSI’s Manila International Container Terminal has implemented the YardSight AI from the AiCON suite for its RTG operation

Industry background

The sister companies are IGO Solutions and Avlino Inc. Headed by director Jay Pandya and Rajan Ramanujam, IGO Solutions is a consultancy with two decades of experience in business consultancy, business process mapping, software development and testing, and automation in container terminals. Led by Ramana Jampala, New Jersey-based Avlino is an AI company founded by serial entrepreneurs Gauthaman Vasudevan and Maryann Sciallo. The firm has developed industry-leading automated reference architecture to specifically deliver industrial AI solutions.

IGO has worked on large-scale greenfield automation and TOS implementation projects, starting with Euromax in 2006, where it performed business process mapping and end-to-end testing for automation systems. Since that time, IGO has completed 12 large automation projects, including for DP World, Abu Dhabi Terminals, Rotterdam World Gateway, ICTSI’s VICT terminal in Melbourne, and most recently Ports of Auckland.

Over the course of that work, IGO has built up considerable expertise and completed business process maps for a large range of terminal processes. It has built this expertise into a range of products that includes automated testing solutions for the TOS, called i-CATS and i-Endurance, the latter of which is specifically for testing releases of Kalmar’s TLS software.

Through its work in terminals, IGO realised the potential of AI to enable terminal operations, and three years ago, it invested in Avlino to deliver AI solutions for the container terminal industry. The company has developed its own ‘Autonomous AI’ engine that uses adaptive machine learning and algorithms to discover patterns in and across large data sets.

In an interview with WorldCargo News, Jampala and Pandya explained that the development took a leap forward when one of IGO’s customers saw the potential for AI tools to be combined with domain expertise in port operations, and allowed IGO access to five years of its operational data. Using these data, Avlino has built AiCON, a comprehensive suite of AI modules for the container terminals that is able to fully integrate into existing enterprise processes and applications.

Getting started

Despite all the interest in and hype around AI, the enterprise AI adoption rate for AI tools is actually quite low. In 2020, the US Census Bureau reported that fewer than 10% of companies in the US were actually using AI. Consultancy firm Gartner has reported that the four biggest obstacles to AI adoption relate to enterprise maturity (not having skilled staff or quality data), a lack of understanding of AI benefits and uses, not knowing how to find a starting point, and concerns about the integration complexity of AI tools on the market.

Pandya said that, in addition to the above factors, terminal operators are wary of starting a complex IT data integration project that will lock them into one AI supplier without really knowing what the outcomes of that project will be.

To address these issues, Avlino developed Alenza, a customisable ‘AI Reference Architecture’, specifically to deliver industrial AI solutions. Alenza brings patent-pending automated AI functionality and adaptive machine learning to multiple data sources. Avlino and IGO solutions worked together to customise the Alenza architecture to deliver AiCON.

The first six modules from AiCON are:

● VesselSight - vessel forecasting and berthing.

● BerthSight - maximising waterside performance.

● YardSight - optimal container stacking.

● PortSight - terminal performance.

● EquipmentSight - equipment failure prediction.

● AlertSight - system health with application failure prediction.

Problem solving

Pandya emphasised that the AiCON modules are all ‘solutions’ designed to dynamically solve a specific problem, rather than predefined rule-based software to manage an aspect of the operation. VesselSight, for example, uses schedule, AIS, vessel traffic and other data to answer the question of when a specific vessel will actually arrive, while BerthSight makes berth allocation decisions and defines load and discharge patterns for the quay cranes accordingly.

Terminals have TOS modules that also make these decisions, but in a much different way. TOS optimisation follows what IGO calls “manually configured engineering rules” that are preset and need to be updated manually. On the other hand, AiCON focuses on what Avlino calls “self-adjusting dynamic rules” to optimise all aspects of the terminal operation, including minimising equipment travel time and distance.

AiCON’s YardSight module, for example, takes real-time data on the current state of the yard, gate, and berth, plus other sources including gate booking systems and historical dwell time data. Pandya explained that AiCON pre-plans the yard for the next three to 14 days to best meet the business rules defined by the terminal in real time. This is particularly relevant today, as yard densities increase with larger call sizes, increasing the amount of digging required to retrieve import containers.

Jampala added: “YardSight continuously learns and adapts, avoiding the necessity of frequent configuration changes. If gate turn times increase, or containers of one particular type start experiencing a very high dwell time, the system can identify the trend and adapt without any input from planners. The process is real-time responsive – taking less than 150-200ms to calculate the optimal decking position for a container.”

Easy integration

A key focus for IGO and Avlino has been to engineer AiCON for simple integration. All the modules are TOS-agnostic and can be connected through a universal API. Terminal operators can subscribe to a module nd start benefiting from ‘augmented decision-making’ that they can input into existing processes very quickly. “There are no changes to the team or workflow,” said Jampala. “We can give you results in eight to 12 weeks. It is a true TOS plugand-play.”

There are, however, still implementation challenges to overcome. Planners and equipment operators given yard positions by an AI tool can still choose to override those instructions and place containers elsewhere, or change an equipment allocation. Pandya said it is important for the terminal to set up a KPI to measure how many times the AI tools are overridden and record the results. There will always be exceptions where AI does not produce the best decisions, but by enabling terminals to compare the outcomes of different decisions, AiCON can help build trust and confidence in AI-based decisions.

IGO/Avlino made a major announcement this month, disclosing that ICTSI’s flagship Manila International Container Terminal (MICT) in the Philippines has implemented the YardSight AI from the AiCON suite for its RTG operation.


MICT operates a fleet of over 49 RTGs and has a throughput capacity of around 3.3M TEU a year. Yard planning in Manila is particularly challenging due to ongoing issues with road truck congestion, customs and very long dwell times in the Philippines.

In an interview with WorldCargo News, Anders Dommestrup, CEO and executive director of MICT, said that AiCON has fundamentally changed the relationship between operations and software at the terminal. Using AI-based software, he said, has enabled the terminal to “break the boundaries with the operation system” and apply its business rules to operations much more directly.

Dommestrup said that replacing predetermined “engineering rules” with regard to speed, acceleration and equipment cycle times with self-adjusting dynamic rules in processes like yard planning is a paradigm shift. Basic averages for cycle times based on machine parameters, he added, are “always inflated, and do not reflect the distribution of cycle times in the real world, where different operators drive differently, and performance varies greatly”.

Real data

AiCON, by contrast, uses real data from the terminal to understand how the equipment really behaves and how long cycles really take. For MICT, Alenza revealed not just the average RTG productivity, but the full distribution of cycle times. MICT found that its RTGs were delivering as little as one move an hour and as many as 35.

MICT wanted to address this with new yard planning rules. Pandya noted that as YardSight does not take a ‘rules-based’ approach to yard planning, it will not produce results that keep equipment operating outside the specified business rules. YardSight does not carve out yard blocks for containers with a fixed set of parameters. Instead, it measures the current state of the yard and then takes into consideration how it will be groomed over the next 30 days. Yard slots are chosen within simple ‘boundary constraints’. For MICT, these included not mixing containers with long and short dwell times, minimising re-handling, and using consignee information as a determinant for dwell time.

Using AiCON’s YardSight, MICT was able to allocate yard moves in a way that narrowed the cycle time distribution, or as Dommestrup put it, “narrow the bell curve”, so the yard work is more evenly spread, and the terminal can achieve a given number of moves with fewer RTGs. Furthermore, MICT is fully in control of the business rules behind yard planning decisions, so it can adjust yard targets if it sees they have an undesirable impact on other areas, such as berth productivity.

Dommestrup highlighted that there is a significant difference in the IT architecture and systems for “optimisation” modules compared with AiCON. Rulesbased planning and optimisation require a lot of integration of different data streams, which is expensive, time-consuming and inflexible. With AiCON, the AI engine itself is a black box, but the business rules themselves are easily coded by terminal personnel.

At MICT, terminal planners, with remote assistance from IGO/Avlino, were able to set up the business rules and start working with AI-based yard planning. According to Dommestrup, the process was no more difficult than a minor update to the TOS. Furthermore, rather than feeling like they are implementing new tools that would replace their work with automation, both operations and IT staff at MICT have embraced AiCON as a tool they can understand and work with.

With regard to results, MICT has seen a significant improvement in work balancing across its RTG fleet, enabling it to better match the number of machines it deploys to the work required. Dommestrup said results include “increased productivity for internal and external trucks, reduced traffic clashes, and minimised yard re-handles when stacking import containers”.

As it moves forward, MICT plans to use AiCON to implement improved business rules in other areas, such as the time taken to connect reefer containers, managing the dynamic allocation of CHE ranges, and achieving efficiencies for the export yard cycle for internal/external trucks.

“We are already seeing very encouraging signs that this new approach can deliver significant efficiencies to equipment utilisation and truck turnaround time, which will be passed on to our customers to further support ICTSI’s conscious effort to reduce its greenhouse gas emissions,” said Dommestrup.

Pandya added that, over time, MICT will build a digital twin of the whole terminal. New opportunities will open up from reviewing each process, without the need to try and tackle everything at once in a long and expensive project.

What makes the journey with AI worth taking is the potentially significant reduction in operational costs that can be achieved through better visibility and planning. IGO/Avlino claims a medium-sized terminal handling 1M TEU per year can expect to improve the productivity of its assets by a factor of 1.3 to 1.5. This, in turn, can reduce its Capex by 30% to 40% over five years.

The automated world

MICT does not have any automated equipment, but Dommestrup believes AiCON’s use of self-adjusting dynamic rules could be equally beneficial to automated terminals. Prior to taking up his role at MICT, Dommestrup was the CEO of VICT in Melbourne, which is a fully automated terminal.

Although terminal automation is making progress, Dommestrup said planning and scheduling are still “guided by the same metrics and rules that are set within the TOS”. Based on the experience at MICT, he sees AI as opening up an opportunity for automated terminals to put business logic back at the forefront of terminal operations.

For a long time, equipment OEMs have been struggling to develop automated systems that can drive equipment with the flexibility and decision-making ability of humans. MICT’s experience with AI for yard planning, Dommestrup concluded, is an eye-opener as to how AI-based autonomous driving combined with self-learning could be a game-changer for the industry.