Investigating human dynamics in small spaces

by Shabnam Rashtchi Manager, Deep Learning (NLP) Lead, PwC United Kingdom

COVID-19 has forced many of us to revisit how and where we work. At the start of the pandemic, few may have considered the use of office space as a pressing concern. However, as we hopefully move towards another easing of restrictions, conducting a safe and efficient return to offices is becoming an increasingly important challenge for many organisations.

The cascade of lockdowns throughout the world forced companies to assess and implement various measures to protect their people. But it soon became apparent that there was neither a clear view of re-opening measures nor a simple way to evaluate the effectiveness of each measure for a specific workspace.

A deluge of publications from the academic community failed to help. These provided too many options, with solutions ranging from installing hand sanitiser stations, to creating dedicated socially distanced queues for the lifts and one-way walk lanes around the buildings. For some, the options were so overwhelming they decided to simply send their people home. While this solution was simple and absolute, it has created a follow-up problem - how and when can people safely return to the office?

Bringing people back is complicated and there are many ways to do it. But one thing is clear: reopening cannot be safe using a single, simple rule: there are a variety of scenarios that need to be tested.

This is where Agent-based Modelling (ABM) can help. ABM is a mature method that enables examination of risky situations in a risk-free environment. However, tools used for this simulation of indoor spaces focus mostly on emergency (e.g. fire evacuation) or crowding analysis. With the pandemic creating a unique set of requirements for combining physical space and human behaviour, a standard ABM approach is unlikely to be effective.

PwC’s AI team has developed an agent-based simulation that provides detailed modelling of how people move around indoor spaces, analysing the likelihood of risky interactions that may increase the chances of virus transmission. This analysis was conducted to measure the effectiveness of intervention measures in keeping people safe in office environments. The types of interventions considered range from social distancing and flow control, to using every other desk and everything in between.

Adaptable to any multi-storey building layout with a large number of employees, the model can identify reductions of space capacity to maintain a safe environment, especially in communal areas such as kitchens, toilets and elevators.

It’s an approach that we’re currently using to assess the reopening scenarios for a number of our offices. Using the unique tools and techniques mentioned above, we hope to perfect the method in the coming months and intend to share our findings.

by Shabnam Rashtchi Manager, Deep Learning (NLP) Lead, PwC United Kingdom

Read more articles on