Technical Independent Study
Saanvi Ravi Katyayan
“Sometimes, it is so crowded I can't even get into the trains”
~ An MTR User
These feedback from everyday MTR users is what made us realize that we need to build a tool for crowd management that can increase the comfort of users, with comfort being defined as minimum congestion and shorter distance = faster movement. This study revolves around the generation of shortest past with minimal congestion
The Spatiotemporal Routing Algorithm
Spatiotemporal Oblivious Routing (SOR) is a routing algorithm designed to provide routes that mitigate congestion in route networks.
Unity Simulation
A simulation is created to mimic the behavior of people entering the Choi Hung station at 6 pm, i.e. a peak time, so that the effect of our robotic barrier and LED guidance can be studied and compared to the results of our practical experiments at HKUST, thereby signifying system effectiveness .
Algorithm key steps overview
Define the Path and Route Network
Where the route network is set containing the starting node, ending node and all the nodes in between in a path along with the travel time from node to node.
Routing query updated
Shortest Path Calculation
Shortest Path Calculation
Calculate Edge Load
LED calculation
Barrier Calculation
When a passenger enters the station, their time of entering is recorded, and a request is made to find the route from specified source to specified destination.
The shortest path is calculated between the nodes specified using Dijkstra's Algorithm
The shortest path is calculated between the nodes specified using Dijkstra's Algorithm
Edge Load describes the amount of people present at an edge (or node) at a given instant of time. The edge load is used as an indicator of congestion status/crowd status for each edge pair.
The LED is set to display a green color for less than 40% of people present in its respective area, orange for between 40%-60% people and a red color for greater than 60% population.
Extended distance to a node is calculated after a barrier is placed in front of the node. Then, the final distance to each node and the passengers initial preferred node is taken into account to determine if the passenger will still go to the same node or change their path after the barrier is placed.
LED Simulation
The model shows the escalators and stairs leading down to the train towards Whampoa at the Choi Hung station. The crowd behavior and layout has been adapted accordingly. At the platform of the Choi Hung station, (lower floor) there are 4 LEDs, each LED is assigned to 10 carriage doors each. When the area covered by one of the LED is overcrowded at the platform, the LED turns red.
The simulation was tested at different percentage of people following the colour shown by LED. It was noticed that when between 37-60% people follow the direction indicated by the LED, none of the areas at the platform remain overcrowded, this agreed with the data from our experiment. In the scene from the simulation indicated above, 40% people follow LED directions.
You will notice that when LED turns red, some people in that area change their colour to blue and move to different areas. These are the people who follow the LED directions.
Barrier Simulation
The model has been revamped to show the effect of the barrier on the top floor of the Choi Hung MTR station. The barrier is placed on the path to one of the escalators, hence increasing the distance a passenger has to travel to visit that escalator. The barrier in front of the escalator is placed in front of the most populated escalator down to Whampoa at the choi hung station. In order to evaluate whether the person who was initially going to use this escalator continues to do so, the new distance with the barrier along with distances to the other routes down are taken into account. In the simulation, you will notice that soe people going towards the barrier turn yellow. This is to indicate the people who change their path once they see the barrier, thereby indicating that the barrier would be effective in a real-life scenario.