top of page

Traffic Analysis Platform

An AI-powered platform to help cities make data-driven decisions for urban design.

This is an Urban Flow project.

A common challenge across our cities is their lack of a data-driven approach towards urban street design. Whether it's conducting traffic counts of various types of vehicles, or redesigning junctions to prioritise the safety of pedestrians and cyclists, cities tend to follow a highly manual, cumbersome and inaccurate data collection process.

This significant shortfall in the urban street design/traffic engineering process is especially apparent in Indian cities. Junctions tend to be unnecessarily large with poorly designed traffic islands, inadequate refuge islands and traffic calming, or have arbitrary geometry that impedes the flow of traffic.

After witnessing these challenges first-hand while working with city planning departments, I decided to build the Traffic Analysis Platform.

Use Cases

With a wide variety of urban design projects being taken up in cities, use cases for the Traffic Analysis Platform are truly endless. From improving the safety & efficiency of junctions, to evaluating the impact of specific interventions and designing pedestrian plazas, this platform is crucial to the success of urban design projects in the cities of today. Conventional methods of conducting traffic studies which tend to be manual & cumbersome, are simply unable to capture the nuances of complex urban environments.

With the safety of pedestrians, cyclists, and vulnerable road users being in focus more than ever before, such a platform is a necessity for designing spaces & streets that are safe, sustainable, and ready for the future.

Frame 25_edited_edited.jpg
trf-ana-sample-metrics.png

How it Works

With a highly diverse set of vehicles present on Indian roads, most existing traffic analysis tools are unable to accurately perform vehicle detection & classification. They also tend to use complicated or expensive methods for collecting data, which may not be scalable in the Indian context. In addition to this, many tend to focus solely on vehicle-centric studies that ignore the needs of pedestrians, cyclists, and other vulnerable road users.

For the Traffic Analysis Platform, we have built a robust dataset of vehicles & pedestrians in diverse street conditions and environments. This dataset was then used to train state-of-the-art (SOTA) object detection AI models for vehicle detection & classification. We also built our own algorithms for accurately counting vehicles, detecting vehicle-pedestrian conflicts, capturing vehicle speeds, and a wide variety of other metrics.

The following process is followed for conducting a traffic study:

  1. Identify regions of interest (RoI) along a street or at a junction

  2. Define study parameters, output metrics, and time slots for capturing data

  3. Temporarily install compact battery-powered cameras on existing street light or signage poles

  4. Capture video footage as defined in the study parameters and time slots

  5. Process video footage with the Traffic Analysis Platform and generate results

This process allows us to conduct traffic analysis projects in an accurate, efficient, and scalable manner.

Output Metrics & Data

Some standard outputs include:

  • Traffic count by direction & vehicle category

  • Natural pedestrian desire paths & refuge islands

  • Vehicle speed heatmap

  • Pedestrian/bicycle conflict zones & crash-risk areas

  • Vehicle travel paths & unutilised road space

  • Temporal distribution of traffic volume by direction & category

  • Pedestrian counting & behaviour analysis

Being an AI-based platform, various other types of output metrics can be customised and configured, depending on the use case and study requirements.

trf-ana-sample-2.png
IMG_2023-01-11-13-13-15-997~2.jpg

Deployment

In January 2023, we conducted a pedestrian traffic study for our client Compartment S4 on Nrupathunga Road in Central Bangalore.

With the objective of building people-friendly street furniture, Compartment S4 wanted to identify regions along this stretch where pedestrians tend to halt for breaks, gather at food stalls, wait for buses & taxis, etc.. Based on results of the study, they would install benches, water dispensers, standing tables, and other amenities.

 

The scope of our study included capturing the following metrics at four locations along the street:

  1. Total pedestrian count

  2. Average hourly pedestrian count

  3. Count of pedestrians walking

  4. Count of pedestrians halting

  5. Walk count by direction

  6. Temporal analysis of walking

  7. Temporal analysis of halting

  8. Scatter plot analysis of halting (volume, duration, time)

  9. Heat map visualisation of halting in the region of interest (RoI)

Analysis Results

Following completion of the study, multiple visualisations and data points were captured allowing unique insights into the behaviour of pedestrians along Nrupathunga Road. These insights will now be shared with city authorities to install street furniture - creating small but critical spaces of refuge & comfort for pedestrians amongst a bustling and fast-paced city.

tr-ana-sample-output.png
bottom of page