While at SXSW Interactive this past weekend, autonomous vehicles proved to be a hot topic with many panels discussing the future and societal implications of autonomous vehicles. Various car companies and universities showed up to give predictions and answer questions about how a world with autonomous vehicles will operate. It’s clear we can visualize a future with autonomous vehicles within the next decade – but how will we make it happen? In my Future15 talk at SXSW 2015, I highlighted emerging hardware technologies that will enable us to lessen the occurrence of car accidents, parking shortages and traffic jams.
Figure 1. Three hardware technologies work together to sense 360° around the vehicle, with built in redundancy to maximize safety.
Innovation in hardware and software – the building blocks for these technologies – is critical to making these ideas a reality. Researchers are looking at three key technologies for sensing: camera, radar and “light detection and ranging” (LIDAR - see figure). Many consumers have already seen back-up cameras available in newer cars, including features that alert the drivers when there is an obstruction in the camera view. Radar has been appearing in new (especially luxury) vehicles for the past few years for collision and side-panel sensing as an alternative to camera technology. Radar systems send an electromagnetic pulse and makes calculations from the return pulse to determine if an object is dangerously close. It provides an additional layer of safety and can be effective when cameras may be less so – think low-visibility conditions. The third sensing component, LIDAR has been used in prototypes for self-driving cars but current LIDAR models are costly – almost as expensive as purchasing the car itself. Because of this high cost, the technology is likely new to the average consumer.
A LIDAR system will send out a pulse of infrared light using a laser diode, detect the reflected signal with a photodetector and compare the incoming and outgoing pulses to calculate the time of flight or difference in time stamps between the two signals. Knowing that light travels at – well, the speed of light – we can determine the distance over which the pulse has traveled. By shooting off thousands of pulses in a well-designed pattern, you can map the surrounding area of a vehicle to distinguish cars, people, trees and other obstacles. Currently, cost is a critical barrier to getting LIDAR in every car on the roads today. My research in Kilby Labs involves examining ways to enhance performance while reducing the costs involved, especially as the costs of the individual components like the laser and detector are limiting. Additionally, we are also working on shrinking the form factor of the system overall, so we can embed multiple devices into the vehicle without affecting aerodynamics, or aesthetics.
Because all three technologies have strengths and weaknesses, researchers are looking closely at how to fuse the technologies together efficiently to make the best use out of the data each sensor can give us. One way to do this is to take all the available sensor data (camera images and video as well as radar and LIDAR information about the surroundings), send it to a central processing brain and then make decisions about how to affect steering, braking and other controls. While that is one method, researchers are actually now looking at distributing the processing power to the sensor nodes and performing some of the calculations at the sensor nodes. From here, extracted information, like an object’s distance from the vehicle or how fast it’s going, is sent to the central processor. This enables the processing costs to go down by reducing the amount of data sent to the central computer. Sensor fusion not only reduces system costs and increases efficiency but also adds an additional layer of safety through checks and balances and redundancy in information.
We’re well on our way to making autonomous vehicles a reality, but there’s a lot of work to do on both the hardware and software sides to bring costs down enough to integrate them into your next new car. Watch for a lot of exciting semiconductor innovations in the years to come, changing the way we think about driving and our commutes.