New Report Shows Sensing Tech Still Lacking in Driverless Car Prototypes

Home > News > Content

【Summary】Out of the three types of sensors available on driverless vehicles, LIDAR seems to be the most effective in managing environmental elements.

  Michael Cheng  ·  Aug 10, 2017 2:05 PM PT
author: Michael Cheng   

Humans are highly efficient drivers. At the moment, driverless cars are still not as capable as people, when it comes to understanding the surrounding area and performing complex maneuvers on public roads. A bit of snow and even darkness could throw off an autonomous vehicle's sensors, resulting in unforeseen risks.

A new study from the University of Michigan's Sustainable Worldwide Transportation is currently tracking the efficiency rates of self-driving systems. Researchers are patiently waiting for the time when smart cars will be more competent than humans behind the wheel. But based on the results of the latest report, it seems that developers still have a long way to go before this achievement can be unlocked.

"People can drive within lanes, even if the markings are faded or disappear altogether. They can cleanly brake for cats and roll through plastic bags that look vaguely like cats," explained Aarian Marshall from Wired.

Human Driving Efficiency

Scientists overseeing the study claim that human senses are better than existing LIDAR and radar components found on self-driving cars. During the trial, humans received "good" ratings for tasks related to object classification, edge detection and lane tracking. Radar units were given "poor" ratings for all three tasks, while LIDAR received "poor" for lane tracking and camera sensors were given "poor" for weather performance.

Human participants weren't provided "good" ratings across the board. In particular, the subjects were not efficient at driving in dark locations. Moreover, individuals received "poor" ratings for interactions with city infrastructure networks. Out of the three types of sensors available on driverless vehicles, LIDAR seems to be the most effective in managing environmental elements. Radar received the most "poor" ratings, but did not have any "fair" ratings. Cameras on the vehicle only received one "poor" rating and a total of three "fair" grades.

V2V Solutions

Unlike humans, autonomous cars can communicate with each other to better understand their environment. Many developers see V2V as the main solution for improving predictions during operation (via Dedicated Short-Range Communication [DSRC] protocols). For congested cities, V2V is suitable, as urban driving has increased 33 percent between 2000 and 2016, according to a report (Recent Diverging Trends in the Amount of Urban and Rural Driving in the United States) from the University of Michigan Transportation Research Institute.

Rural locations are experiencing less drivers on public roads, at a decreased rate of 12 percent over the same period, due to more people attempting to move closer to big, busy cities. Out of all the different types of vehicles, light trucks make up most of the units on the road today. Private automobiles come in second, next to heavy, commercial trucks.

"To really have the best possible integration of an autonomous vehicle you need to have the connected DSRC involved so the vehicles can talk to each other instead of sort of having to see and sense their way through the world," said Brandon Schoettle, project manager at Sustainable Worldwide Transportation who authored the study, during an interview with Jalopnik.

Prev                  Next
Writer's other posts
    . Related Content