Modelling crowd behaviour for ultra-large ship design

Ultra-large cruise ships are designed to accommodate 7000 people so layout design is crucial to optimising passenger experience as well as safe evacuation procedures. In a one-year Cohesion project, ME researchers Austin Kana and Bilge Atasoy, of the Department of Maritime and Transport Technology (MTT), joined forces with Javier Alonso-Mora at the Department of Cognitive Robotics (CoR) to model how passengers move around super large cruise ships with the aim of improving ship layout design.

“MTT is highly interdisciplinary,” says Atasoy, “so even though we work in the same department, Austin and I are doing very different types of work. I look at transportation problems, optimising how things move, and I’m particularly interested in how behaviour plays a role there. For example, why does someone choose to send a container via one route as opposed to another? If we understand this behavioural aspect, we can design better transport systems.”

“And my background is in ship design, primarily early stage design for complex ships such as cruise ships, naval vessels and ships that service wind turbines,” explains Kana. “And as part of that, we need to make decisions about ship layout - where to put what types of technology - and that was what was interesting about Bilge’s research - modelling how people make these sorts of decisions in different contexts.”

We wanted to look at ship design from the perspective of how the crowd moves in a ship so we could design a better layout; for instance, design the aisles differently or plan better locations for different facilities.

How robots move around humans

Kana met Atasoy while working on a project that involved simulating evacuation strategies on large cruise ships, and they discussed setting up a Cohesion project together. Atasoy: “We wanted to look at ship design from the perspective of how the crowd moves in a ship so we could design a better layout; for instance, design the aisles differently or plan better locations for different facilities.”

Kana also approached Alonso-Mora at the Department of Cognitive Robotics: “I just “cold-called” Javier and asked him about his experience with robot-modelling and movements.” And Alonso-Mora was happy to come on board: “We have a lot of experience in Reactive Collision Avoidance, which is robots moving around humans, so that fits very well with analysing how crowds move.”

Wuhan and Covid-19

The Cohesion team were also keen to make use of the expertise of Yapeng Li, a guest researcher who had spent at year at MTT working with Kana: “But in November 2019, just before the project started, Yapeng had to go home to Wuhan in China - and the rest is history!” explains Kana, recalling the early days of the Covid-19 pandemic with Wuhan at its epicentre. Fortunately Atasoy knew other people who were working on choice models including PhD graduate, Jishnu Sreekantan Nair at the Department of Civil Engineering (CE). “Although Jishnu didn’t know much about ship design, working with someone from a totally different background worked really well,” says Kana. Nevertheless, there were a few delays since Sreekantan Nair had to build his own simulator, and that took time: “This meant that by the end of the project, he hadn’t got as many results as he had wanted.”

Both safety and passenger experience

Understanding human behaviour is all-important when designing an evacuation procedure for a large and complex ship, especially when there are 7000 people, all behaving differently: “Many standard models make the assumption that you just follow the most direct way out,” says Kana, “but in reality, you’re going to be influenced by things like your mobility - children move around differently from elderly people or people in wheelchairs. Or if you’re with friends and family, you might start looking for them first.”

“And there’s also the influence of the behaviour of the crowds around you,” continues Atasoy. But evacuation procedures was just one part of this project: “We started looking at how people generally move about in airports or hospitals, taking inspiration from big buildings and architecture, then applying that to large cruise ships. So these were our two aims - to make these ultra-large ships safer but also better from a passenger enjoyment perspective.”

We want to make these ultra-large ships safer but also better from a passenger enjoyment perspective.

Bilge Atasoy

Researcher

Choice model with a difference

A year later - and despite the holdups as a result of the pandemic - the team has a good basis for further work. Atasoy: “We have a range of results and we can include different possibilities about how people make choices - and we have a basic choice model which is a little different from what’s out there in the literature.” So with both the simulation and the data ready to go, a little more work should see the first paper being published soon. “If I’m honest, I think one year is a bit too short for a Cohesion project,” reflects Kana. “It takes a few months to get up and going – and once you’re running, you’ve only got two or three months to finish and if we’d had a few more months, I have no doubt that we’d have something beautiful by now.”

Freedom and good brains

Meanwhile their Cohesion project has brought other benefits. Alonso-Mora: “What I’ve learnt, other than getting to know Austin and Bilge, is that it was nice to see the different ways to model behaviour and to model how people move – and we can use that to make better robots as well.”

As for Cohesion projects in general, Atasoy is enthusiastic: “One of the important things is that we are not tied to any particular industry – we have the freedom to do fundamental science, try crazy ideas! So I think that’s a positive thing about Cohesion – maybe you don’t always reach the final goal but at least you brainstorm about it - and you brainstorm with good brains!”