AARG

Our research focuses on developing Multi Agent guidance systems to control and manage crowd. For this we have considered the scenario of the Tawaf ritual at Hajj, Mecca.
To understand it, We performed a downscaled Crowd simulation of Hajj using various models such as Social force model, CA model, ORCA, etc using the Menge Framework. We analysed crowd density in different concentric circular regions using Poisson distribution to control the Inter-region rates of movement. We found interesting results suggesting the ways in which people tend to move closer to Ka’aba. This causes high crowd density as one moves closer to the center. It becomes increasingly difficult to manage crowd.
We are working on an approach that uses Multi agents to control the Crowd to prevent congestion. The Goal is to maximize pilgrim crowd flow and to ensure no stampede (High density regions). Previously researchers have proposed some trajectories, structural amendments that reduce the risk of stampede. Our approach is novel and dynamic. We wish to design Multi Agent system to guide groups of people to perform circumambulation and to touch the Kaaba. Multi agents schedule movement of groups to minimize the time to complete ritual. We form groups of people, each group controlled by a single agent. To find the Optimal Policy we formulate it as an MultiOperation Optimization problem. Given flow rates of Multi agents(Poisson distribution: lambda, Mu)
Constraints/Tasks for each agent:

• Complete N full rotations around center.
• Reach the center at least once for prayer.
• Stay at the center for approximately T time( Can be later optimized )
• Center can hold maximum of C agents at any given instant(Analogous to praying near Kaaba for some fixed time)

Optimize: Minimize the total time delay for each agent, indirectly increase agent flow.
Plot shows the crowd region density for each ring. We can see that first the inner ring fills up then the middle ring and then finally the outer ring. The average time between arrival of any person is 1( lambda) and that of leaving is 1.1(Mu). We model it using Poisson distribution and follow the rule that a person given a choice always tends to move towards the inner ring until their is no space left.