CIOInsider India Magazine

Around 1,50,000 people are killed annually in road accidents in India, as per government data. Road rage, increase in vehicles, poor transportation infrastructure, skipping traffic signals, not wearing helmets or fastening belts, using mobile phones, drunk driving, sudden malfunction of vehicles, half-done construction work, non-usage of construction signs, and the list goes on. Even the college and school roads that need immediate repair are left unattended for months and years, which throws a big question on whether the country is heading towards a better and safer future.

Moreover, there is immense interest in implementing electric vehicles (EVs), with a certain amount of them already hitting the streets. It’s important to ensure proper transportation infrastructure to support and encourage the usage of EVs. The more EVs ride in, the closer we are to a zero emission future.

Many promises are only uttered through the mouth and neglected by the hands. How many deaths before someone or something take a step?

Today, the government is employing artificial intelligence (AI) to make roads in India a safer place to drive. With safety comes a reduction in road accidents. The plan is already being executed in Nagpur, Maharashtra, called ‘Intelligent Solutions for Road Safety through Technology and Engineering' (iRASTE). Let’s deep dive into the technological framework that makes it do what it does.

It doesn’t End with AI; There’s More
Of course, AI is deployed in the project and helps identify the cause of the accident while driving a vehicle. Once that’s discovered, the next action it takes is alerting the driver before the accident occurs with the help of an Advanced Driver Assistance System (ADAS).

The project is headed by I-Hub Foundation, IIIT Hyderabad, a Technology Innovation Hub set up in the technology vertical- Data Banks and Data Services, and the Department of Science and Technology (DST) is providing support.

What’s more, the I-Hub Foundation, having a history of serving the latest technologies across data-based technological solutions in the mobility sector, is deploying machine learning (ML), computer vision, and computational sensing. The proof is the India Driving Dataset (IDD), a data set developed for road scenes for understanding unstructured environments, using the worldwide assumptions of well-delineated infrastructures, such as lanes, limited traffic participants, and low variation in the object or background appearance, and strong adherence to traffic rules.

Now let’s debunk how these technologies are alerting drivers.

Identifying the Grey Spots
The unique AI approach undertaken in the project helps predict as well as identify risks on the road.

Firstly, it is about identifying the grey spots. This is done with the assistance of both data analysis and mobility analysis that never take their eyes off spotting dynamic risks that could occur in the entire road network. Identifying the grey spots is a top priority since undressing them could turn them into black spots, and that’s bad news. Black spots mean showing locations with fatal accidents.

While both systems don’t divert their eyes from the entire road network, it helps them come up with ideas such as engineering fixes and even design the same to correct existing road blockage for preventive maintenance and improved road infrastructure.

There’s more. The Technology Innovation Hub is working towards coordinating, integrating, and amplifying basic & applied research across broad data-driven technologies. It is also working on its dissemination and translation across the country.

Where is the Project Headed?
The AI in the project will help prepare a critical resource for future use by researchers, startups, and related industries with a keen focus on areas such as smart mobility, healthcare, and smart buildings.

Additionally, AI and other technologies will help devise a blueprint with practical solutions that suit Indian conditions. For now, it is being rolled out in Nagpur; eventually, it will make its way to other cities across the country. Currently, there’s much interest shown by the Telangana Government to adopt the technology for its bus fleet that takes the highways.

Similarly, other projects such as Open World Object Detection on Road Scenes (ORDER), Mobility Car Data Platform (MCDP), and LaneRoadNet (LRNet) have been appointed for the job, with each manifesting unique methods to ensure road safety.

What is the ORDER?
To understand ORDER, it is first important to know that object detection is a key component in autonomous navigation systems enabling localization and classifying objects on the road scene. Take existing object detection methods, which are trained and inferred on a fixed number of known objects present in road scenes.

Whereas in the real-world, many unknown objects exist that were not witnessed by the system during its training. Therefore, the government’s ORDER is specifically designed to tackle this issue with the help of Feature-Mix, which improves the unknown object detection capabilities of an object detector. Next, the system identifies the road scene dataset by comparing it with a generic object dataset that consists of important parts of smaller objects. With higher intra-class bounding box scale variations, detecting known and unknown objects becomes challenging.

To counter that challenge, curriculum learning is deployed, and finally, an extensive evaluation of two road scene datasets is provided.

On to MCDP
With several sensors such as cameras and LIDARS, along with necessary computing, it makes it easy for any individual to capture and process data on the car. Also, it serves an educational and experimental purpose for researchers and startups to test their automotive algorithms and approaches in the navigation and research on Indian roads.

LRNet is Deep Learning the Same but Differently
With an integrated mechanism that looks into lane and road parameters, this system deploys deep learning to learn as well as address the problems with Indian roads. This system carries out a road quality score which is calculated with the help of a modular scoring function. The final score helps authorities to assess road quality and prioritize maintenance schedules on the road for improved drivability.