I Got Early Access to Waymo’s Self-Driving Taxis. Here is what I learned.
If you've been in Los Angeles in recent months, you've likely seen one of Waymo's fully autonomous vehicles driving the roads. Waymo, formerly Google's driverless taxi arm, has been developing autonomous vehicle technology since 2015 and currently operates in Phoenix and San Francisco, with limited access offered in Los Angeles. When I was back home in LA for winter break, Waymo offered me a week of early access to its fleet of self-driving cars, an offer I was eager to take.
Currently in its 5th generation, Waymo's fleet is composed of electric Jaguar I-Pace SUVs retrofitted with cameras and sensors that make the vehicles easily identifiable from the outside. Waymo uses 29 cameras mounted on the sides and top of the car, six radar sensors tracking larger objects and four LiDAR sensors, which use laser pulses to "see" the car's surroundings in 3D. LiDAR provides a highly accurate scan of the car’s surroundings, even at night, since it doesn't depend on an external light source. While most automobile manufacturers offer factory-equipped assistive driving technology that relies primarily on cameras and radar, these pale in comparison to LiDAR sensors and are generally considered inadequate for fully autonomous vehicles. Elon Musk, a notable exception, has maintained the feasibility of fully autonomous driving technology that relies only on vision-based sensors.
Using Waymo is similar to requesting a ride on any standard rideshare app. You enter your pickup and dropoff location and any waypoints on the Waymo app, and the company dispatches a vehicle. The app allows you to select an exact location on the map for the car to stop (if safe) and even lets you choose the side of the street you'd like picked up on. The average wait time of around 15 minutes was longer than my usual experience with Uber or Lyft. When the vehicle arrives, you unlock it through the app and get inside. The fleet of Jaguar I-Paces currently seats four since the driver's seat is off-limits. Stickers warn you to keep your hands off the steering wheel as "the Waymo driver is in control at all times."
Once inside, riders can start the ride via the app or the two infotainment screens in the interior. Throughout the trip, these screens augment what Waymo "sees" through its sensors, giving riders a greater sense of confidence in the machinery. They display a modeled image of everything from lane markings and surrounding vehicles to pedestrians and traffic signs. Most importantly, Waymo highlights its intended trajectory with a green line, so you can be sure it isn't planning on driving into that truck in front of you.
After only a few miles with the Waymo driver in control, I began to trust it implicitly. The company's track record definitely helps here. According to the NHTSA, Waymo has driven a total of 7.14 million miles and beat human benchmarks with a 57% reduction in police-reported crashes and an 85% reduction in crashes involving injuries. In my experience, the vehicle was also able to react to various irregular road conditions. The Waymo driver identified traffic cones, maneuvered through construction zones and pulled to the side of the road to allow emergency vehicles to pass. The car also slowed for speed bumps and sped up to catch yellow lights. I was most impressed when the car maneuvered around a stopped tow truck blocking multiple traffic lanes.
One of the most noticeable aspects of the experience is the reaction Waymo elicits from other road users. It is not uncommon for people to pull out their phone at a stoplight to record Waymo in action or for pedestrians to stop dead in awe. There seems to be a lot of interest in the technology, and people often stopped me to chat about the car when exiting. Nowhere did Waymo get more attention than the carpool line at my grandmother's senior daycare center, especially among the older people who were waiting to be picked up in their "normal" taxis.
Waymo has gone a long way to make the experience comfortable for riders. You can control the air conditioning from the back seats and play music from the infotainment screen. However, it still has room to improve the passenger experience to gain widespread adoption. At times, the autonomous system can feel robotic. When making an unprotected left turn, it can be hesitant or indecisive, with a series of sharp starts and stops before completing the maneuver. Steering inputs may also become abrupt, particularly when lane markings change suddenly. Waymo also doesn't seem able to control the vehicle's chassis. It cannot identify and avoid bumps and potholes in the road and tends to choose the (less comfortable) rightmost lane to drive in. Furthermore, Waymo currently refrains from driving on freeways with passengers aboard, which could significantly increase travel times for longer trips in Los Angeles.
During my 150-mile experience, Waymo encountered a significant issue only once. It got stuck on a single-lane road because it couldn't communicate with another driver. In that instance, the system contacted customer service to take over.
Despite these shortcomings, I never felt unsafe inside Waymo. If anything, the system appeared to err on the side of excessive caution, even if it compromised on comfort. Having long touted the ability of self-driving cars to reduce traffic accidents, Waymo conducted a study where they simulated fatal crashes and researchers substituted the initiator (the car that caused the collision) and the responder (the affected car) with the Waymo driver. For example, in an instance where a human driver runs a red light, the autonomous system would take the place of both the at-fault and affected drivers. In the initiator role, Waymo avoided 100% of collisions. Perhaps even more impressive, Waymo anticipated the human error in the responder mode and took corrective action, avoiding 82% of crashes and otherwise reducing the severity. The remarkable results demonstrate that Waymo can perceive and react to challenging situations better than human drivers.
Overall, I found my experience with Waymo to be impressive. As Waymo's technology advances, incorporating more human-like qualities, such as refining the Waymo driver's inputs and establishing a system for effective communication with other road users, could significantly enhance rider comfort and increase adoption. If Waymo can address the relatively high cost of self-driving hardware, it stands to gain economic advantages, such as operating a 24/7 fleet and realizing savings in employment costs. This could position Waymo competitively compared to other rideshare services in the market of the future.
___
Photo Caption: Waymo’s autonomous taxi driving on city street in San Francisco.
Photo Credit: Dietmar Rabich / Wikimedia Commons
