No Driver, No Problem?
Google hosted an update on its self-driving car (SDC) project in early October that included ride-alongs in two test mules. Motor Trend podcaster-in-chief Charlie Vogelheim and I ventured to Google headquarters in Mountain View, California, for the event, which started with rides and concluded with a presentation by the SDC team.
Our first ride was in a modified Lexus RX450h, one of three apparently identical SDCs Google made available for our junket. Google started with Toyota Priuses when it began experimenting with self-driving technologies, and the Lexus SUVs appear to host Google’s latest self-driving hardware and software alongside traditional driver controls, namely a steering wheel and gas and brake pedals. The redundancy is logical enough; when testing autonomous vehicle technology on public roads, it will be wise to have a driver and controls available should something go wrong.
Our Lexus came with two co-drivers, one behind the wheel, the other in the front passenger seat, monitoring a laptop, so Charlie and I rode in the second row. Aside from a short but wide-aspect ratio monitor mounted high on the center console (more on that later), there wasn’t much visibly different inside the Lexus, except for a large round, red button mounted next to the shifter—ostensibly to be hit in emergencies. Despite the light modifications, Google requested no interior photos of any of the vehicles in which we rode.
The exterior was more heavily modified (and photographable), most notably with an ice-bucket-sized spinning array atop a roof-mounted rack. “The Lexus has a plethora of sensors hung around the perimeter of the vehicle,” quipped Charlie at first sight. “The vehicle is as much a test bed of sensors as a display of autonomous capabilities.” To sense its environment, Google self-driving cars rely on camera, radar, laser (LIDAR), and global positioning (GPS) systems mounted at various positions on the vehicle. Cameras are generally used for monitoring short to medium range distances and evaluating conditions such as changing traffic signals, objects in the immediate vicinity, and parking situations. Radar can be used in short-, medium-, and long-distance applications. Because it uses radio waves, radar is largely unaffected by weather conditions. LIDAR, the spinning getup on the roof, reflects laser beams in a 360-degree field of view, giving the onboard computers an accurate representation of the surrounding environment, and GPS provides precise location information. How Google’s computers and software stitch all of this data together is a discussion for another time, but the actual guidance of the vehicle is executed by the Lexus’ electric power steering, throttle, and brake systems.
So how does all this, uh, ride?
Well, despite how chaotic it may sound in the Motor Trend podcast Charlie and I attempted to record, I found our roughly 12-minute journey interesting, if a little underwhelming.
After gawking, then boarding and buckling ourselves in, we were off—slowly at first, and then to a full stop, our longest delay of the day as we waited to merge to the right out of the Google parking lot. It was a somewhat busy street with the roadway to the left obstructed by a curve in the road, so our Googled Lexus made plenty sure that the roadway was clear before pulling out.
Once underway, it was business as usual. The drivers told us that the top speed of the self-driving RX450h is 35 mph (56 km/h) and that the vehicles are programmed to obey all the traffic laws and posted signs. That seemed to be the case as we puttered around the suburban neighborhoods that surround the Googleplex HQ. Charlie and I did note that at one point the speedometer indicated that we were doing more than 25 mph (40 km/h) in a residential area that shifted to a school zone.
“Good to know that it would only go through a school zone at 15 err, 25 err, 28 mph (45 km/h),” Charlie said. Our drivers had no comment.
As we cruised along neatly trimmed neighborhoods dotted with mid-century Eichler homes, there was time to evaluate that short and wide multicolor display mounted high on the center of the dash. We’re bombarded by screens in cars these days, from Tesla’s giant touchscreen tablet-style display to Audi‘s screen-only virtual cockpit, but the slim yet spare screen in the Google Lexus is surprisingly refreshing. The road ahead glowed green against a dark field that occasionally rendered oncoming roads and objects in purplish-pink. Trees and cars and other obstacles we approached were regularly rendered every few seconds in whitish squiggles. The pulsing manner in which the environment was drawn, apparently by the LIDAR unit spinning above our heads, was reminiscent of scenes from old World War II submarine movies, where the sweep of the radar reveals the position of nearby ships and terrain. The road ahead and simple rendering of the surrounding area were just about all the information the screen conveyed during our trip: no speeds or traffic data, little in the way of street names or signage. Occasionally a “yield” notification would pop up, as if to explain why the vehicle was slowing.
There is precious little to report otherwise. The ride and handling were smooth, quiet, unremarkable; turns out a Lexus at 35 mph (56 km/h) is a Lexus at 35 mph (56 km/h) no matter who is (or isn’t) driving. With Google software at the wheel, there were no abrupt stops or changes of direction, no jerky or unsmooth driving that would indicate anything other than a human chauffeur was behind the wheel. Charlie did have mixed feelings about the way the lane changes were initiated, “The [auditory] lane change announcements were both informative and intrusive,” he said. “Imagine doing the same as you were driving.”
The only other distractions to the self-driving car experience: a bit of fan noise and heat emanating from computer hardware hidden behind the rear seats—and the human co-drivers. They were both friendly enough and happy to answer our rapid-fire and no doubt annoyingly basic questions, but their mere presence took a little of the autonomous magic away. While Charlie noted how odd it was to see the steering wheel move by itself as the vehicle negotiated corners and intersections, I found it equally strange to watch the human minder, hands hovering around the steering wheel, feet covering the pedals, ready to step in should the vehicle veer toward imminent disaster.
Is that really necessary? I thought to myself. Some of this blas attitude is, no doubt, a product of my own familiarity with the current state of semi-autonomous driving systems. A week before this Google ride, at our annual Car of the Year testing, we marveled at the new Chevy Malibu‘s adaptive cruise control and lane keep assist systems’ ability to hold an 80-mph (129 km/h) cruising speed around a 6-mile oval while the driver took his hands off the wheel for the entire ride. Sure, asking a vehicle to maintain speed and lane position on a closed highway is less complicated and risky than asking a car to drive itself around a crowded Silicon Valley suburb, but Google should take comfort; the trust in its technology and others is out there.
Or perhaps I’m speaking too soon and this is but false bravado. Is a true driverless car something to fear? We soon found out.
Sure, the Google SDC looks cute in a dopey sort of way, but that’s by designâa nonthreatening and approachable design dictated by safety. Make no mistake, what Google is trying to do by creating its own self-driving car is very different from what has made it a leader in the technology space. A bad search result, a glitch in your free email account, smartphone operating system, or goofy web-connected glasses might give you a headache, but an error in a self-driving car could do much, much worse.
With such thoughts swirling, Charlie and I embraced the sinister goofiness, opened the door, and climbed into the small electric-powered two-seater.
“Very welcoming, plenty of room,” Charlie said as he shut the door. For this portion of the ride-along, the Google Self Driving Car team had journalists staged atop one of the Googleplex parking garages, queuing for two-minute rides along a predetermined course. “Kind of a plasticky, mass-transit feel to the interior fit, finish, and switches.”
That’s by design, too. The Google team worked with a number of well-known auto industry suppliers to build the prototypes and early versions of the car, but plans for consumer sales have yet to be revealed. Instead, Google hinted that its first application might be in some sort of shared ride service as a way to expose as many people to the new technology as possible.
That would explain the paucity and simplicity of the controls; between the two seats are familiar switches for the windows, seat heaters, and reading lights and a glowing “Go” button that toggles to”Pull Over” once depressed. Behind it lies a red-trimmed button under a plastic shield, presumably for emergency situations.
In front of both passengers is a large and deep bin for holding bags and gear; it has an edge that doubles as a footrest. On the dash above the bin was a screen similar to the one found in the Lexus. Behind that is the car’s large windshield, which we later learned is made from flexible plastic to protect pedestrians from impact. On the A-pillars, two rectangular screens show views from the front corners of the car. Not sure if they will be on the car when it heads into production, as they seemed very prototype-ish. Also curious was a vertical and horizontal slit in the footwell area. It seems the ideal spot for some sort of pedal, perhaps to pop out in an emergency? Who knows.
After belting in and handing over our podcast tools (microphones and handheld recorder), we were ready to go. In the real world, you’d push the button and tell Google where you’d like to go, perhaps via your phone (the team is still working this out), but since this was a preprogrammed route, we just pushed the Go button and slowly rolled out.
The Google team set up a simple route complete with obstacles meant to simulate a few real-world situations. After negotiating a 180-degree U-turn into a straightaway, the vehicle slowed for a pedestrian (a Google employee) who abruptly crossed our path. We negotiated a couple more turns and yielded again as another Googler pulled out in front of us in a Ford Fusion (also white, also a hybrid—notice a theme here?). After slowing and safely passing, we yielded again for a bicyclist who approached from the side. As he pedaled toward us, the dash-mounted screen continued to pulse and render his form, traffic cones, and structures around us. Fun to see what the car “sees.”
After rounding the last bend, we sidled up to some cones and slowed to a stop at a predetermined spot a few paces behind where the next riders were to be picked up.
“The ride was benign, sterile, very similar to an airport tram or Disney ride that calmly accelerates to a predetermined low speed,” Charlie said after the ride. “We’ve all been on the ride a hundred times, yet this time there wasn’t a rail. The evasive moves and stops were unremarkable and predictable. I would like to have experienced a panic stop.”
I wished for more, as well. Google’s self-driving technology is impressive; our short rides in its two cars clearly demonstrated that there is plenty more capability beyond what the team showed us. But modesty is understandable given what is at stake. Google is the highest-profile non-car company developing autonomous driving technology and considered by many to be the smartest guys in the room. Any mishap in the early rollout to the press and public would represent a setback not just for Google but for the entire emerging category. Still, there are moments when you just can’t help but be impressed.
“The most compelling moment was when we got out and the empty car proceeded forward to the next rider,” Charlie said. “Something about the vehicle moving without an occupant drove home the autonomous character.”