Radar was deemed as a necessary safety feature in autonomous vehicles until Tesla announced they will not be shipped with radar moving forward. Dealing with inclement weather like fogs and night is a difficult problem for LiDar and cameras.
What is the problem with radar? A single radar system suffers from poor imaging quality in that when only a small fraction of radio waves are bounced off objects, causing unknown. A Multi-radar system might improve the perception with 2 sensors 1.5m apart but still cannot guarantee accuracy. The hardware and complexity to merge vision from cameras with radars increased the cost of the car.
What is the problem with using a single radar system?
It receives just a few points to represent the scene, so the perception is poor. WaveSense president and co-founder Tarik Bolat, ground-penetrating radar could be on autonomous cars circa 2024 with a cost to produce on the order of $100 in quantity. (extremetech.com)
The WaveSense project was tested in the military and scanned 10 feet below the roadway surface to get a unique identifier that is accurate to an inch or two. Mapping cars would scan the roadways once, then your self-driving car with its own ground-penetrating radar would rescan as you drive, matching its real-time scan to the master map. That would keep your car centered, even if pavement markings are covered by snow or ice.
What are the benefits of multi-radar systems?
“So if a single radar is causing this blindness, a multi-radar setup will improve perception by increasing the number of points that are reflected back.”
A team at the University of California San Diego found that spacing two radar sensors 1.5 meters apart on the hood of the car was the optimal arrangement. The team “hid” another vehicle using a fog machine and their system accurately predicted its 3D geometry. (therobotreport.com)
When the team "hid" another vehicle using a fog machine, their system accurately predicted its 3D geometry while the LiDAR sensor essentially failed the test. (wkyc.com) Waymo, the main supporter and building their LiDAR sensors, has also managed to cut the price of LiDAR sensors by almost 90% in recent years. (fierceelectronics.com).
What are the results of the multi-radar system?
The team developed new algorithms that can fuse the information from two different radar sensors together and produce a new image with less noise. The system worked well when tested at night and in foggy conditions.
The problem is the dataset is really small at 54,000 radar frames and inconclusive at the moment (imagine the number of million miles Waymo and Tesla have in comparison).
Elon Musk famously said on 10 April 2021 that vision has much more precision and better to double down on vision than do sensor fusion with radar.
What is the dataset?
“We collected our own data and built our own dataset for training our algorithms and for testing.” The dataset consists of 54,000 radar frames of driving scenes during the day and night in live traffic, and simulated fog conditions.
“There are currently no publicly available datasets with this kind of data, from multiple radars with an overlapping field of view,” Bharadia said.
What's next for self-driving cars?
A new kind of radar could make it possible for self-driving cars to navigate safely in bad weather.
This way, we don't need to use expensive LiDARs, which Tesla or OpenPilot had been touting. The system consists of two radar sensors placed on the hood and spaced an average car's width apart (1.5 meters).
As mentioned, WaveSense has built a radar system to scan what's below the road, down where there's no snow at all, rather than parse wintry mix on top. This requires complete knowledge of the world map which is also not viable in the long run.
What are the risks of driving in bad weather?
The car's sensors can be blocked by snow, ice, or torrential downpours, and their ability to "read" road signs and markings can be impaired.
In Boston, where NuTonomy has been road-testing autonomous vehicles in cooperation with city planning officials, snow and seagulls have emerged as two of the biggest obstacles. It is unlikely someone will train their dataset against seagulls without the actual data.
Just like with any new technology, autonomous cars have faced their fair share of speed bumps from concept to product, especially when it comes to navigating in inclement weather.
Conclusion
Overall, there are many radar development as cars are still unable to achieve full autonomy with current "vision" technology. it will be exciting to see if which technology cameras with ultrasound, LiDAR, or even radars will rule the market.