If you have panicked on the thought forced to Parallel Park, you are not alone.
There are many of us who have a problem with the vehicle maneuver.
But, there is technology now that can help us do it easily.
Jayant Thakur, Director of the software processor in Texas Instruments, said his team in India had spent the last five years developing the Surround Scene technology.
“There is a camera around the car, and when you try to park or drive on a narrow lane, the camera provides sewn views to help the driver,” he said.
Recently, the team has used machine learning and in-depth learning to increase this surround view technology with strong new capabilities.
“Now, you just have to go to the parking lot, get out of the car, press a few buttons and let the car park itself,” said Thakur.
Texas Instruments provides system-on-chip (SOC), basic software, deep learning features, and parking examples for some of the largest Tier-1 car makers and original equipment manufacturers (OEM) in the world.
The company then helps car makers and OEM products and customize technology for their individual vehicles, because each vehicle has a unique speed, handle, rotating radius, etc.
“We optimize the SOC capabilities to get the highest performance on the lowest power consumption.
This technology will eventually attract from the car battery,” said Thakur.
Piyal Goswami, which is part of the Thakur team and leads the side of software development from parking technology, said the team must deal with a large number of data from various sensors.
“From the processing perspective, there are two parts – perceptions and navigation.
Our main challenge is that we have many sensors, such as radar and lidar, whose raw data must be massaged so we can extract some useful information from it.
That’s because we need to detect in Certain frames whether there are pedestrians walking in front of the car, or other vehicles, or even distinguishing between trees, lines, driving rooms, etc., “he said.
The pictures are distinguished by ML assistance, especially neural network models.
“The main challenge for the software team is to make different models learn all different angle case scenarios to reduce the error rate,” said Goswami.
Automatic parking is a more difficult challenge in India because of the unstructured environmental nature such as a large pole or stone in an unexpected place.
“Getting the data model so that the car maker and OEM can capture all corner cases where the car can hit India while parking is quite a challenge.
But we hope to solve unstructured environmental problems,” Goswami said.