Even as people hear about exciting developments regarding autonomous cars, many have lingering doubts about their safety. Videos of these vehicles show them manoeuvring around obstacles and dealing with other traffic without incidents, though.
By Kayla Matthews
That's possible due to high-tech sensors that provide advanced detection capabilities. They're on other cars that aren't fully autonomous too, such as those that recognise impending collisions and alert drivers to use the brakes.
These sensing technologies are already impressively advanced, but there's an ongoing push to make them better. Today's autonomous cars have cameras and radar or lidar sensors that work together and make a vehicle apply the brakes a mere half-second following the detection of an obstacle.
That's still faster than the average human driver response time of 1.6 seconds. However, a car fitted with the camera and sensors travelling at 30 mph would move for more than 20 feet after the brake application.
Researchers at the Fraunhofer Institute for Reliability and Microintegration IZM believed a better method existed, and they partnered with a team of industry leaders and fellow researchers to find it.
A reaction time of under ten milliseconds
The improved radar sensor the team came up with has a much faster reaction time than the ones currently available for cars. This one responds in fewer than ten milliseconds, making it 50 times speedier than others on the market and 160 times quicker than a human driver. A vehicle equipped with the sensor also progresses for only about 6 inches after the brakes are applied.
The researchers think their technology could help stop inner-city accidents. Those could be especially dangerous, considering the potential for extreme congestion and how the impact on one car could affect others nearby.
What makes this system different?
The researchers' achievement is a module with cameras on either side. The outside looks like a gray box, and it's about the size of a smartphone. The unit stands out because all the signal processing required by the device takes place in an integrated manner. Moreover, the onboard technology filters the input based on whether it originates from the car's camera or the radar.
The system also recognises irrelevant data and does not designate it for evaluation. Then, the processing of the information can begin immediately or start after a certain time period based on the current settings.
Next, sensor fusion technology combines the data from both sources, and neural networks use machine learning to analyse the information and its real-world traffic implications. This process results in the sensors only sending instructions for the vehicle to react to obstacles rather than feeding it status information, too.
This technology could make autonomous cars safer by significantly reducing the time between when a vehicle detects something in its path and when it stops.
The need for testing
If this system eventually makes it onto the majority of autonomous cars, it must pass tests first. Since 70% of the components of a vehicle come from the supplier chain, testing starts before manufacturers install parts onto the car. Mechanical and environmental tests also determine whether a finished vehicle can withstand the conditions associated with everyday drives.
However, autonomous car testing will not be as straightforward as it may seem to some outsiders. These cars represent entirely new territory regarding what technology can do. Regulators will not have previous benchmarks they can use to determine what makes an autonomous car "safe enough" for widespread use.
Moral dilemmas could also arise, such as if a car must either go into the path of pedestrians on a sidewalk or veer into a light pole, which should it do? The former could keep the passenger safer, but it could harm more people.
Improving heavy traffic conditions
The work mentioned above about the improved sensor is not the only research occurring regarding how to make autonomous cars perform better in the presence of other vehicles or obstacles. Another study, this one from the University of Cambridge, looked at how a fleet of driverless cars worked in tandem to improve traffic flow.
The researchers discovered that autonomous vehicles could perform in tandem to positively influence traffic flow by as much as 35%. They did this by sending signals to each other that changed the actions of all the cars in the fleet. They could even avoid a human-driven car operated aggressively.
What's next for the enhanced radar project?
Concerning the better radar sensors for cars, the scientists involved will keep working on the project until it ends in 2020. From now until then, some of the project's partners will continue putting the prototype through its paces — including launching a Berlin road test. It's too soon to say for sure how this improvement could make self-driving cars better, but the prospects seem bright.