• New Depth Sensors Could Help Self-Driving Cars
  • January 23, 2018
  • Researchers who are part of the Camera Culture group at the Massachusetts Institute of Technology’s Media Lab have developed a computational method that significantly improves the resolution of time-of-flight depth sensors. Time-of-flight measures the amount of time it takes for light that is projected into a scene to bounce back to a sensor. Their approach, which increases the depth resolution 1,000-fold, has the potential to make self-driving cars more practical.

    For the technique to be effective, there must be good detectors. The MIT researchers used interferometry, which involves splitting a light beam in two. Half of the light beam circulates locally, while the sample beam is fired into a visual scene. When the two beams are recombined, it produces an exact measure of the distance that the sample beam travelled. This approach is much more practical in self-driving vehicles.

    According to Ramesh Raskar, Associate Professor of Media Arts and head of the Camera Culture group, the fusion of optical coherence and electronic coherence is a new concept. By modulating the light at a few gigahertz, it is comparable to turning a flashlight on and off millions of times per second. Rather than controlling light optically, we are changing it electronically, said Raskar. The combination of the two results in power for the system.

    Overcoming the Problem of Fog
    In addition to the improved accuracy, new research suggests that the method can accurately measure distance through fog. This has been an issue that has created problems for self-driving cars. Fog can be problematic for many reasons, but for time-of-flight systems specifically, it scatters light by deflecting the returning light. This can cause light to arrive late and at odd angles. It can be extremely difficult to isolate a true signal with interference.

    Today’s time-of-flight systems have a depth resolution of approximately one centimeter at a range of two meters, which is fine for assisted-parking and collision-detection systems that are equipped in newer cars. However, according to a PhD student, the MIT system was able to achieve the same resolution at a longer distance. At a range of 500 meters, they were able to reach a depth resolution of a centimeter. This can help self-driving cars detect objects in the road sooner, avoiding serious accidents.

    Baltimore Car Accident Lawyers at LeViness, Tolzman & Hamilton Protect the Rights of Self-Driving Car Accident Victims
    If you or a loved one has been in a car accident, it is in your best interest to contact the highly skilled and experienced Baltimore car accident lawyers at LeViness, Tolzman & Hamilton at your earliest convenience. Like any new technology, manufacturers of self-driving cars are still perfecting certain aspects of their vehicles and the accident risk is largely unknown. Our dedicated team will pursue the maximum financial compensation you deserve and ensure that your rights are protected. To schedule a free consultation, call us today at 800-547-4LAW (4529) or contact us online.

    Our offices are located in Baltimore, Columbia, Glen Burnie, and Towson, allowing us to represent injured victims in Maryland, including those in Anne Arundel County, Carroll County, Harford County, Howard County, Montgomery County, Maryland’s Western Counties, Southern Maryland, and the Eastern Shore, as well as the communities of Catonsville, Essex, Halethorpe, Middle River, Rosedale, Gwynn Oak, Brooklandville, Dundalk, Pikesville, Nottingham, Windsor Mill, Lutherville, Timonium, Sparrows Point, Ridgewood, and Elkridge.