How Autonomous Vehicles Handle Austin’s Thousands of E-scooters
There are several effective ways to get around downtown Austin, Texas: drive your car or hail one, ride a public transportation bus or the Capital MetroRail (a light overground rail system that connects the downtown area with Austin’s suburbs), or, increasingly, use an app on your smartphone to unlock one of the many dockless e-scooters or e-bicycles that have proliferated around the city’s streets over the last four years.
“I feel like every day there’s a new company that has scooters on every corner,” says Martin Berger, a Test Specialist (TS) for Argo AI, a self-driving technology company conducting testing in Austin.
Berger’s job, along with the other Argo TS’s, entails riding in the front seats of autonomous test vehicles — Ford Escape Hybrid SUVs and Ford Fusion Hybrid sedans — to evaluate the performance of the Argo self-driving system.
The TS job is critically important, as Argo prepares to provide an autonomous ride-hailing service to passengers in Austin through Lyft beginning in early 2022.
In order to provide a safe and comfortable service, Argo’s TS’s monitor how the company’s self-driving system navigates around Austin day and night, and how it handles encounters with scooters — including ones that may not be following the rules of the road.
Startups have deployed scooters on many city streets as part of a growing movement to enable more micromobility options but Austin stands out for the sheer number. As of late 2021, there are at least 11,600 rentable scooters and 500 shared bicycles from four companies, according to the city government’s website.
If you walk around downtown day or night, you’ll see people whizzing down the streets on scooters at speeds of up to 15 miles-per-hour, and packs of them parked in clusters in neon green, purple, silver and white.
Riders find them convenient to hop on and go, but drivers may struggle to see the plethora of small, quick vehicles or to anticipate some of their maneuvers.
Given that’s the case, how do autonomous vehicles fare when contending with Austin’s scooters?
The Argo self-driving system “does handle it really well” when encountering scooters, even those riding erratically or in violation of the rules, says Maya Zbeda, an Argo TS in Austin.
That’s in part because Argo’s self-driving system is designed to use the multiplicity of sensors placed precisely around the car’s exterior — including cameras, radar and laser sensors known as lidar — to detect objects both stationary and mobile with a high degree of precision, accurately measure their speed and direction, and forecast their future motion and trajectory.
In particular, the Argo self-driving software has two classification methods: a “closed world” classification system that, when it receives enough data to identify an object with a high degree of certainty, classifies it as a pedestrian, animal, vehicle, or other “wheeled object” such as a bicycle or scooter; and an “open-world” system that detects unusual objects that can’t immediately be classified into a precise category — say, a custom parade float — but which the car nonetheless detects, measures and takes proper driving actions to safely navigate around.
Pedestrians, bicyclists, scooter riders and other small objects are all considered “vulnerable road users,” or VRUs, by the Argo self-driving system, because they lack the protection afforded by a larger vehicle body, such as a car or truck. Argo’s self-driving system therefore places them into a special category that makes the system drive with extra caution around them.
Argo’s self-driving system is also adept at handling all kinds of road users, even those deviating from posted rules, whether they are cars traveling slightly over the posted speed limit, pedestrians crossing the street outside of crosswalks, or scooter riders blowing through traffic lights.
The same is true of Argo TS’s, who are trained to react safely, calmly and wisely around everyone on the roads during their four-week certification course before riding in the test vehicles.
“I have zero faith that a scooter or bicycle is going to stop at a red light,” says Cleophus Joyner, another Argo TS in Austin. “So you have to be extra cognizant of them.”
Joyner notes that the University of Texas campus is located in downtown Austin, and when school is in session, the streets become filled with students walking and scootering between their dorms and classes, presenting an especially dense and often chaotic driving environment.
“The toughest area [to drive] is around campus,” Joyner says, both for humans and autonomous vehicles alike.
That’s coming after the campus enforced an 8-mile-per-hour speed limit for motorized scooters using geofencing technology – wherein scooter companies detect where their vehicles are through global positioning and use their onboard software to limit the top speed accordingly.
During their test rides, Zbeda and other Argo TS’s will disengage Argo self-driving cars from autonomous mode and drive manually when required to by their training, as well as anytime they feel it would be helpful for the flow of traffic and safe to do so.
For instance, as outlined in Argo’s safety report released in early 2021, TS’s are “currently required to take manual control when encountering an active school bus, or when they encounter a first responder vehicle with emergency lights flashing.” This will be the protocol until Argo soon completes integration and testing of features that will allow its self-driving system to safely handle these situations.
Another part of the TS job is to review on a computer how the Argo self-driving system would have performed on its own in the occasions when TS’s did disengage autonomous mode and take over driving, through a feature known as playforward – a simulation tool that enables Argo to analyze scenarios that occurred on the road.
Zbeda says the Argo self-driving system drives safely around scooters both in real life and in playforward. Even the times when she has performed autonomous disengagements around scooters, “it’s really cool then to go back then through the playforward and see that the car would have stopped.”
Overall, Zbeda, Joyner and the other Argo TS’s who spend their days riding around Austin feel that Argo’s self-driving system is well equipped to handle the range of scooter behaviors around Austin.
“The [autonomous vehicle] always impresses me,” Joyner says.