Ai Shall Soon Monitor Driving Tests.
Did you ever hear any Indian Driver’s License holder say that they had to work extremely hard, or even moderately so, to get their hands on their license? We have posed the same question to a bunch of people and as we expected, almost every time, the answer was a resounding no. Every Indian major holding a driver’s license knows the level of simplicity with which licenses are handed out. Whether it is a learners’ license or a permanent one,
getting hold of either of these is a piece of cake for Artificial Intelligence.
Now moving on to the after math of these unmonitored handouts, reckless driving is one of the main causes of accidents in India. Everyday, if you flip through the newspaper you will notice reports on at least one gory accident. Whether it is a resultant of speeding or jumping the signal or even a simple error in indication, dire consequences are faced by drivers on the road.
70.4% of the accidents caused due to Over Speeding
6.3% of the accidents caused due to Wrong Side Driving
1.4% of the accidents caused due to Jumping the Signal
Now, if you are wondering, just like we did, how many of these accidents were even caused
by license holders, here are some stats that will stop you from giving the traffic system the
benefit of doubt.
Accidents caused by License holders 79.9%
Accidents caused by Non-License Holders 10.4%
Well, with the government issuing stricter rules to discipline the existing traffic, we are sure that a change is on order. But this isn’t the only measure that the government has decided to take up to spread awareness about road responsibility. Issuance of licenses is soon going to be a highly monitored process with Artificial Intelligence taking over this responsibility.
HAMS, Harnessing Automobiles for Safety, is the brainchild of a research team at Microsoft, that had begun to take shape in the year 2016, a decade after the idea was first attempted to conceptualize. The revolution of Artificial Intelligence combined with the rising usage of Internet of Things has enabled this project towards successful completion.
This smart-phone based driving test system was initially being used to train drivers at the Institute of Driving and Traffic Research (IDTR). But, as anticipated and planned earlier, the artificial intelligence product has now been promoted to work as the sole monitor of driving test candidates. In the first-of-its-kind case, HAMS has been installed in the test vehicles at the Regional Transport Office(RTO) of Dehradun and has successfully taken place of a
human evaluator and it is also perceived to have the ability to work with much higher competency. Higher competency most definitely sounds like great news, but aren’t you curious about the technicalities that are providing HAMS its competence?
Alright, then let’s dig in!
Researchers Akshay Nambi and Venkat Padmanabhan of Microsoft Research India have conducted research on various aspects for the development of this safety system. Below, we have jotted down all of these research aspects that have been embedded into HAMS as individual features and also the techniques and concepts that each of these aspects are built on. Take a look!
FarSight is the function that enables to find the distance of the test vehicle from the preceding vehicle, in other words, vehicle ranging. The function uses basic trigonometry to calculate the distance from the vehicle using a bounding box after identifying the class, or say the category, of the vehicle. The visuals are recorded by the smartphone’s rear camera. Though Radar and Lidar technologies can also perform vehicle ranging, they are not
available in older vehicles.
The DeepLane function allows the system to detect the lane position of the driver. Moving away from GPS systems due to inaccuracy, the DeepLane feature uses deep learning techniques to achieve computer vision to accurately identify the lane position of the test vehicle even in the absence of lane markings. Researchers have proudly announced that this feature displays a 90% accuracy. Well, we would definitely call this a job well done.
The FullStop function has been enabled to detect unprecedented stopping of the vehicle. Again, similar to FarSight, the system uses the rear camera of the smartphone to gather the visuals to detect unsafe stops. Researchers deem this feature to be highly useful in training bus drivers.
The AutoRate function rates the attentiveness of the driver. Spatio-temporal feature Learning is the concept that is used for behavioral analysis and this concept has been applied here. Several spatio-temporal features have been combined to attain higher levels of accuracy.
5. Summing it all up with ALT | Automated Driver License Testing
As we have already mentioned earlier, HAMS was first used for training purposes and was later optimized for monitoring test candidates. The Automated Driver License Testing is the final research aspect that enables the system to be well optimized to enable handing in of licenses only to qualified drivers.
This has enabled the windshield-mounted smartphone to test the test candidate on various metrics such as the following:
i. The front facing camera of the smartphone detects whether the authenticity of the candidate through facial recognition, whether the candidate has his seat belt on and also whether or not the driver scans the mirrors prior to taking a turn or shifting the lane. ALT uses novel auto calibration method based on multiple-view geometry
which helps in calibrating the shift of driver’s gaze in a particular direction. The auto calibration has proved itself to be much more robust than manual
ii. The rear camera of the smartphone is used to detect the driver’s skill in various driving aspects such as parallel parking and driving around a round-
about. To enable this, trajectory estimation is to be done and for this purpose the researchers have developed a novel hybrid Simultaneous Localization and Mapping(SLAM) technique to overcome the limitations of accuracy in a conventional SLAM technique.
The supervision of the driver has been enabled using a fleet management dashboard. On a side note, we would like to mention how well the system has been designed to be accommodated for Indian roads. As you may have noticed above, the system works well for all kinds of road and traffic management.
As we have already mentioned, the system is up and running in the RTO of Dehradun and according to a recent statement given by Mr Nambi, of the 50 odd test takers of Dehradun, only 50% of them are being issued licenses which proves to us that this in fact can improve the traffic conditions by making the test qualifications tougher. This makes us wonder if HAMS combined with the Government’s stricter traffic policies might just make Indian roads a safer place to be on. With artificial intelligence experts leaving no stone unturned to prove that technology is not a danger to the mankind by coming up with more and more socially beneficiary innovations and developments, more and more people from among the audience are slowly jumping on board the futuristic train. This makes us wonder whether technology and humans can indeed peacefully co-exist in the upcoming automated and artificially intelligent future.
On this note, we hope we have given you something that is worth knowing and this is where we bid adieu.
A Computer Science graduate by education and a content writer by profession. Currently fulfilling her zeal to write by putting pen to paper every time she comes across something that is interesting enough to let the world know