SP Company & Workforce Transformation (SP CWT)

PSAM_Banner

 

INSPIRING SUCCESS STORY

PSA Marine: Tapping AI to boost navigation safety

Key Highlights:

  • One of the world’s first AI applications in maritime simulator training and assessment
  • Slashes manual review times for simulator exercises from up to an hour to just 5 minutes
  • Enables training gaps to be plugged quickly with objective, data-driven evaluations

 

Enterprise Profile

PSA Marine is a wholly owned subsidiary of PSA International, a leading global port operator which began as the Port of Singapore Authority in 1964.

PSA Marine plays a vital role in Singapore’s growth as the world’s busiest container transhipment hub. Over the years, its portfolio has expanded beyond towage and pilotage to other businesses such as marine advisory, navigational audit, transportation, training, and consultancy services.

With a 1,800-strong workforce and a modern fleet of more than 80 tugs, it has established a presence in more than 10 countries across Asia, Europe, the Middle East, and the Americas.

Problem statement

Despite the growing use of technology to aid seafarers in navigation, human error remains a key contributor to marine accidents – and insufficient or inadequate training is often the main cause of such lapses.

Besides technical skills, seafarers are also trained in and assessed on non-technical skills such as situational awareness, decision-making, and communication to enhance safety and efficiency. As such evaluations are usually conducted by experts and seasoned instructors, human bias or variability is unavoidable.

An alternative solution was thus sought to ensure consistency and impartiality.


The AI-based Competency Assessment Tool for Safe Navigation (AICATSAN) will ensure consistency and impartiality in training and evaluation.

 

Impact achieved

In September 2021, the Centre of Excellence in Maritime Safety (CEMS) in Singapore Polytechnic embarked on a 30-month project with PSA Marine to design an evaluation framework that leverages the latest technology.

Called the Artificial Intelligence-based Competency Assessment Tool for Safe Navigation (AICATSAN), it aims to offer objective evaluation during the training and assessment of PSA Marine harbour pilots.

By integrating AI with its training programmes and assessments, PSA Marine can ensure that its pilots are well-trained and prepared for demanding situations, said Mr Jimmy Koh, PSA Marine’s Head of Pilotage and Digital Transformation, who leads a team of managers and 280 harbour pilots to deliver pilotage services to vessels that call at the Port of Singapore.

“AICATSAN also enables PSA Marine to identify the strengths and weaknesses of its harbour pilots, and design personalised training programmes to better achieve the desired performance and learning outcomes while reducing human error,” he added.

The first phase of the project wrapped up in March 2024 and the second phase, which will take 12 months, is expected to kick off in July 2024 to further enhance AICATSAN’s features. When it is rolled out fully, the assessment tool will offer three key benefits:


  • Higher efficiency: The integration of advanced sensors and AI algorithms will slash the time needed for evaluations. Typically, assessors monitor and scrutinise trainee performance through video playback of training exercises on navigation simulators. These manual reviews usually take 45 minutes to an hour. In contrast, the AI-based tool can generate comprehensive reports within five minutes.

  • Improved accuracy: AI allows for a more accurate and comprehensive evaluation of critical soft skills such as situational awareness, stress management, and communication – skills that are challenging to measure through human observation.

  • Better personalisation: By providing more objective and data-driven evaluations of employees’ competencies, AICATSAN helps identify areas of improvement. PSA Marine can then tailor or reinforce training to plug any gaps and enhance pilots’ responses to demanding scenarios.

Besides harbour pilots, AICATSAN’s use case can be expanded “to include ship navigating officers so that more seafarers can benefit”, said Mr Koh. As it is, the Singapore Maritime Academy is poised to adapt the AI technology developed in this project, and several shipping companies have expressed interest in using AICATSAN for crew assessment purposes.

 

The new tool will prepare PSA Marine harbour pilots well for any demanding situation.

 

“AICATSAN also enables PSA Marine to identify the strengths and weaknesses of its harbour pilots, and design personalised training programmes to better achieve the desired performance and learning outcomes while reducing human error,” Mr Jimmy Koh

 

CEMS' perspective

The project was a complex one that entailed the design of experiments, creating and selecting simulation scenarios, as well as data collection using various sensors such as eye trackers and microphones.

This first phase also involved the development, fine-tuning, and validation of AI algorithms; data analysis and interpretation; and the establishment of a Competence Assessment Framework, which spells out the standards and criteria for safe navigation to facilitate training interventions.

The entire process required CEMS and PSA Marine to work hand in hand, said Mr Daniel Zhang, Centre Director of CEMS.

The team from CEMS included a human factors researcher who designed the experiments and interpreted the results; AI and software engineers who developed AI algorithms and user interfaces; and simulator technologists who managed data collection.

Meanwhile, experienced instructors from PSA Marine selected simulation exercises and provided expertise on competence evaluation, and harbour pilots took part as subjects to aid in data collection.

“This collaboration ensured a holistic approach to the development and validation of the training tool,” said Mr Zhang.

The team overcame several challenges to enhance various aspects of accuracy as they experimented with different sensor combinations, AI algorithms, and data processing methods to optimise the tool's performance.

For instance, the accuracy in speech recognition was impeded by variations in pronunciation and accents, as seafarers come from many different countries. The team then tested various deep learning models and spelling correction strategies to cut the word error rate in speech recognition to around 10 per cent – lower than the required accuracy of less than 20 per cent.

The application of AI in maritime simulator training and assessment is one of the first of its kind globally, and was presented in 2023 at the International Maritime Organization in London.

“This achievement strengthens Singapore's leadership position in this area,” said Mr Zhang, adding: “It was great working with PSA Marine to spearhead and demonstrate innovation in the maritime sector, which is traditionally viewed as a conservative industry.”

 

RELATED READING

13licenses_thumb

Inspiring Success Story

AMTC: Recovering precious materials from e-waste in greener ways

Discover how new chemical formulations by AMTC help plants that recycle e-waste extract raw components in a safer, more cost-efficient manner.

Read More >

Irvins_Thumb2

Inspiring Success Story

Local snack brand goes global

Working with the Food Innovation & Resource Centre in the mid-2010s has enabled Irvins to expand operations and export its signature salted egg snacks worldwide.

Read More >

logo