Here's the ironic thing about today's autonomous car development programs: The latest prototypes actually require that the person behind the wheel concentrate more, not less.
That's because while self-driving vehicles like this Nissan Leaf Piloted Drive 1.0 prototype can do a remarkable job negotiating roads on their own most of the time, a drive of any length and complexity almost always carries with it the specter of an occasional flub or near miss. By contrast, were a human behind the wheel, most of these situations would have never escalated to the point where a need for a momentary swerve or panic braking resulted.
That may sound discouraging, but it's not meant to. The radical progress that has been made on autonomous vehicles in just the last couple of years suggests that such incidents will be nothing but a brief transitional hiccup for the technology, a blip on its evolutionary timeline. In fact, my 40-plus minute Nissan test drive in unrestricted, live Tokyo traffic was nothing short of hugely impressive.
Even with the current state-of-the-art tech's momentary autonomous foibles, it's easy to see the promise such vehicles have for greatly decreasing accident rates and traffic congestion, not to mention for restoring autonomy to the world's elderly and infirm. Autonomous technology isn't just a game-changer for personal transportation, it's poised to usher in a whole new game.
On my drive, Nissan's all-electric hatchback prototype executed a complex drive route including merging, along with left and right turns. It negotiated dense traffic including busses, commercial trucks and pedestrians, all with minimal intervention. It was truly fascinating to have a front-row seat while the car moved itself nearly seamlessly through dense traffic.
Were it not for the logos slathered on the sides of our, our fellow motorists would likely have never suspected that this was anything other than an ordinary human-piloted car. That's a remarkable achievement. Just a couple of years ago, a car with half of this vehicle's capabilities would've had the outward appearance of a science project.
In part, this dramatic progress has been made possible because the Leaf's complex network of cutting-edge cameras, sonar hardware and lidar sensors (remote-sensing technology which uses lasers and radars to measure distances) have been miniaturized and innocuously mounted to its bodywork. Up until now, these sensors have been large and ungainly, incorporating an attention-grabbing, aerodynamics-spoiling spinning element that needed to be mounted on a vehicle's highest point (the roof) for a 360-degree view.
This Nissan makes use of groundbreaking pre-production flash lidar sensors from Santa Barbara's Advanced Scientific Concepts Inc. which are exponentially smaller than rooftop sensors and contain no moving parts. I bet you didn't notice them subtly flush-mounted on the car's front doors and in the bumpers in the pictures above. In all, this car has no fewer than a dozen cameras, four lidar scanners and five radar sensors attached to its panels. I bet you didn't notice those, either.
Two years ago, a previous-generation system utilized just five cameras and employed comparatively bulky and primitive laser sensors. In the case of this Leaf, which is one of three such million-dollar prototypes, this network of sensors is acted upon by a trunk full of wires and silicon chippery. This hardware figures to be much easier to miniaturize than the sensor arrays themselves, and Nissan expects to downscale the associated componentry to the size of a laptop by the time it introduces a production system by 2020.
At the wheel keeping tabs on the prototype's systems during my test ride was Tetsuya Iijima, Nissan's general manager of its advanced driver assistance systems (ADAS) and autonomous engineering department. Iijima has been working on ADAS and self-driving technology at Nissan for a remarkable 18 years -- he spearheaded the development of the company's first intelligent cruise control, which he proudly notes was among the very first systems in production over a dozen years ago (concurrently entering the market with a similar technology from Mercedes-Benz).
During our drive, Iijima constantly scanned a trio of screens, including a 10-inch center-stack display that looked production ready (the current Leaf features a 7-inch unit) and a 12-inch screen in place of traditional analogue gauges. There's also a third, temporary-mounted tablet screen mounted just ahead of the round gear selector displaying additional information, and a head-up display. The in-cluster display is the one Iijima spent the most time looking at. It shows a radar image of the traffic ahead highlighting objects the sensors detect with either red or green boxes depending on their relevance. These objects can include other vehicles, traffic lights or pedestrians.
The vast majority of our test loop was uneventful in the best possible way -- the Leaf accelerated, braked and negotiated turned with alacrity under almost all circumstances. It signaled nearly every time it changed lanes, and it even functioned perfectly in a long tunnel with a substantial bend in the middle and a fair amount of traffic.
After noting our smooth progress, I asked Iijima whether the instant-on torque of electric motors and the lack of a multi-speed transmission makes developing an autonomous EV easier than a traditional internal-combustion vehicle, and his face lit up as he nodded.
Given that there are occasions where Piloted Drive may require human intervention, the person in the driver's seat needs to continue to pay attention to his or her surroundings. Even when this technology is perfected and autonomous drive becomes at least as safe as human piloting, Iijima says it will be important for drivers to remain awake and maintain situational awareness. To that end, he says, "We may need something to keep the driver involved in the driving task." That could take the form of vehicle occupant monitoring, something this car doesn't do (but Nissan is working on), or it could even take the form of something like an interactive game. This, too, may be a temporary need. In its just-revealedat the Tokyo Motor Show, Nissan envisions a future where the steering wheel actually folds away when not in use, and where the seats shift toward each other to encourage conversation.
As it turns out, fast-approaching cars or an oblivious pedestrian at a crosswalk seems easier for the Piloted Drive to react to than a nearby vehicle whose speed closely matches that of our car -- it's hard for the sensors to determine relative speed in such scenarios.
To illustrate the point, while queuing in slow-moving traffic, our Leaf attempted to merge into the rear end of a commercial truck, coming within inches of wiping off its own front end in what would've been a sickeningly expensive slow-motion accident. Only Iijima's override with the brake pedal saved our car's skin. Piloted Drive also had trouble detecting a merging Nissan X-Trail under similar circumstances. In that situation, Iijima was able to use the steering wheel as a momentary override, at which point the car immediately retook control seamlessly and uneventfully. Piloted Drive won't disable itself fully in the latter circumstance, it will only turn off when the driver manually activates the brake pedal or hits an emergency kill switch -- integrating momentary manual steering or accelerator inputs seem easy.
Interestingly, while Piloted Drive won't call out lane-change maneuvers or bends in the road, when it comes time for a significant directional change (as when turning on to another road at a four-way stop), the system will call out its maneuver over the speakers, just like a normal navigation system. Iijima says that's so that vehicle occupants aren't taken by surprise. IIjima believes that voice support like this could be a transitional step until people get used to the sensation of traveling in autonomous cars.
The auto industry's pace of development of autonomous technology has been nothing short of thrilling, and much of the basic "blocking and tackling" work is nearing production readiness. But there's still lot of scenarios for which it will be very tough to program. For instance, while Nissan's Piloted Drive can account for a pedestrian in a roadway, it can't detect that the individual is actually a policeman, urging the vehicle's driver to proceed with a wave of his hand.
Weather remains a challenge, too -- Iijima says sun glare and heavy rain isn't a problem, but admits fog and snow are more difficult. Will an autonomous car's exterior sensors need to be heated and self-cleaning in order to work in inclement weather, or will the technology simply shut off when sensors get dirty or packed with snow, the way today's intelligent cruise-control systems warn check out?
Nissan promises a full-autonomy system like this will be in production by 2020 and Iijima says it will have a "common-sense" price, but admits it will be initially offered on a high-end model. By that time, less capable self-driving systems will also be available on more mainstream Nissans, which will likely be hardware similar to the semi-autonomous tech currently offered on luxury cars like theor Nissan's own .
In the face of such impressive technology, I couldn't help but ask Iijima if he thinks manual driving has a long-term future, especially if self-driving cars ultimately prove to be significantly safer, as most experts believe. Iijima pointed to the continued presence of motorcycles in a world where four-wheeled cars are infinitely less dangerous and suggested that "(driving) will become kind of a sport," in other words, a source of entertainment. Many experts agree, pointing to how some people enjoy horseback riding even though the sun has long since set on using equines as primary transportation.
Me? I'm not so sure. I welcome autonomous technology, especially for boring freeway transits, megacity stoplight-to-stoplight slogs, and for drivers who can't be bothered to look up from their mobile phones. But I suspect that over the long haul, if human beings are causing a disproportionate amount of accidents versus their computerized counterparts, manual driving will be taken off the table, at least on public roads. Whether that reality manifests itself legislatively or merely practically -- likely taking the form of prohibitively high insurance rates -- I'm not so sure.
Either way, while the road to autonomy isn't fully mapped out, it's clear we're well on the way to a self-driving future. And thanks to engineers like Nissan's Iijima, that future is coming up in the rearview mirror far quicker than you might imagine.