This spring, a high-tech, self-driving car traveled across the country, carrying its passengers safely from coast to coast with very little intervention from the otherwise unoccupied "driver." The car, designed by Delphi, set off from the Golden Gate Bridge on March 22, and drove 3,500 miles to New York.
The trip, the longest autonomous drive cross-country in a car, was accomplished with 99 percent of the drive in fully automated mode over nine days, the company said. Two teams of three Delphi engineers were in the vehicle (one behind the wheel, two in the backseat analyzing sensor data), each team riding along for half the drive.
The coast-to-coast trip allowed Delphi engineers to capture important technical data and road test new safety technology. According to the company, the team collected nearly three terabytes of data.
"We've tested the car in crazy Las Vegas traffic and on highways in California, so we decided to take an extended trip to see all of the different scenarios you can run into," says Delphi CTO Jeff Owens. "We wanted to take that information and use it to make the sensors better and more cost effective, and to guide forward development."
The nine-day trip crossed 15 states and the District of Columbia. Along the way, the vehicle encountered complex driving situations such as traffic circles, construction zones, bridges, tunnels, aggressive drivers and a variety of weather conditions.
Delphi used an Audi SQ5 modified with new radar, laser scanner, and camera technology. Another version of the car (an Audi A7) was previously demoed in Las Vegas at CES 2015 and driven roughly 560 miles.
While a few states have passed legislation to regulate the use of self-driving vehicles, Owens says that the company didn't have to make any special accommodations to operate the vehicle from state to state, other than having a driver behind the wheel at all times.
"We also had to follow the speed limit exactly," Owens says. "You have to program the vehicle at the speed limit, because it would certainly be hard to explain why you programmed it to go above the speed limit. But that meant we got passed by everybody. We weren't just the first to go coast-to-coast in an automated vehicle, we were probably also the first car to do it going the speed limit."
Sensor-packed car
The trip puts Delphi ahead of other autonomous vehicle projects launched by the likes of Google, Apple and other better-known tech companies. The Delphi vehicle also looks relatively normal, compared to earlier versions of Google's self-driving cars, which were equipped with bulky, roof-mounted equipment.
"Our vehicle looks uniquely normal," Owens says. "All of the sensors are styled in, and to the untrained eye it looks like a normal vehicle. You can also open the trunk and put your luggage inside it. The other prototypes I've seen have a trunk full of laptops and power supplies."
Among the technologies in the Delphi car:
• Radar, vision and Advanced Drive Assistance Systems (ADAS).
• Multi-domain controller: High-end microprocessor to seamlessly drive multiple features and functions.
• V2V/V2X: Wireless vehicle communication technology extends the range of existing ADAS functionality.
• Intelligent software that enables the vehicle to make complex, human-like decisions for real-world automated driving.
• Traffic jam assist.
• Automated highway pilot with lane change (on-ramp to off-ramp highway pilot);
• Automated urban pilot.
• Automated parking and valet.
The vehicle (worth approximately $500,000 including all the electronics) is equipped with six LIDAR sensors, six radar sensors, and multiple cameras (including one pointed at the driver to monitor attentiveness).
The car is able to make complex decisions, such as stopping and proceeding at a four-way intersection, timing a highway merge or calculating the safest maneuver around a bicyclist on a city street.
Even with all of that technology, the driver occasionally had to take over. Passing semi-trucks caused the car to move over farther than was necessary, and the vehicle did not automatically move to the left lane when passing emergency vehicles parked on the shoulder of the highway. The "driver" also had to take over in some construction zones.
"You run into construction zones that have quick zig-zag patterns, and in areas where there is no lane definition, the driver would initiate manual control out of an abundance of caution," Owens says.
Variations in lane markings also posed some problems. "The markings are different from state to state, and even among municipalities," Owens says. "That affects how the system understands lane boundaries."
Owens says that one of the key takeaways from the drive is that a combination of radar and a vision system will likely be the way the auto industry will go as automated driving develops. "Vision systems can have problems in a low sun angle, such as at sunset or sunrise, and they have trouble with inclement weather or road mist," Owens says. "Radar had no problem at all."
Owens says that while Delphi continues to analyze the data from the trip, the company will continue to work to lower the cost and improve the performance of its sensor technology. "Next year we'll see some vehicle-to-vehicle systems coming to market from GM, and those will use our technology to talk to other cars," Owens says. "That will provide more information to the vehicle about what's going on, so you can make more intelligent choices about evasive actions."
Owens also expects several of the sub-systems in the autonomous vehicle to continue to expand in new vehicles, even though fully self-driving vehicles are several years away from commercialization. "Long-range and medium-range radar and vision systems are already seeing significant penetration today," Owens says. "We have about 2 million radar units on the road to date. This is easily the fastest growing space in the auto industry, and will be through the end of the decade. We will see tremendous safety benefits because of these technologies."
Subscribe to Aftermarket Business World and receive articles like this every month….absolutely free. Click here.
