The CommutePool service (image: City of Bellevue)

Bellevue is planning a “CommutePool” network of autonomous rideshare services. The goal is to launch the service in 2019, connecting riders from southeast King County to major employers in Bellevue and Kirkland.

Last week, Bellevue and Kirkland jointly submitted a $3 million grant application to the US Department of Transportation’s ATCMTD program (Advanced Transportation and Congestion Management Technologies Deployment). That program funds model deployments of new technologies for transportation safety and efficiency. A decision on the grant is expected by October. The project also anticipates matching funds from both cities and contributions from major Bellevue employers in the Commute Trip Reduction program. Employers already pay for employee use of vans, so this can be an important source of matching funds. The projected cost of launching the program is about $9 million.

Riders will access the system via a smartphone app which Amazon will develop. The app can select pick-up and drop-off locations, find available parking at park-and-rides or leased parking areas, and reserve seats at specific times. Luum, a Seattle firm actively managing commute options at many Eastside firms, will coordinate the program at participating businesses.

The CommutePool service would connect SE King County cities to major employers in Bellevue and Kirkland (image: City of Bellevue)

Bellevue would like to operate the service with autonomous vehicles, but could lease non-automated electric vehicles as needed to start service by mid-2019. An RFP to obtain vehicles is likely soon after the grant is approved. Waymo plans to purchase 82,000 autonomous vehicles for deployment beginning 2019, and General Motors have also said they will deploy autonomous vehicles commercially next year.

In Bellevue, the CommutePool vans will serve major buildings and designated curb pick-up and drop-off locations.

The proposal has buy-in from many of the largest Bellevue and Kirkland employers. Employers with some 33,000 employees on the Eastside have signed letters of commitment to support the program.

The target audience, commuters from cities such as Renton, Kent and Auburn, face long commutes and some of the worst traffic in the region on SR 167 and I-405 south of Bellevue. This market was selected as the initial focus for CommutePool because of its congestion and high SOV share. Development is dispersed so transit service is less available and less convenient in most neighborhoods. CommutePool ridership goals are for 1,000 daily riders within six months of launch; 2,000 after year two; and 3,000 by year three.

Several participating firms already operate their own shuttle services on fixed schedules from leased parking lots. Others subsidize rideshare on Uber and Lyft. This creates both a natural market for the CommutePool service and a funding pool for expansion.

The aggressive schedule places Bellevue in the forefront of cities planning for autonomous vehicles. That incurs some obvious risks that the technology will not be ready, or that the public may not regard commuting in autonomous vehicles as sufficiently safe. Poll data suggests declining confidence in the vehicles, but that may recover as they become increasingly visible. Bellevue expects to use safety drivers in the beginning.

What are the prospects for success? The performance of many micro-transit start-ups has been underwhelming, dogged by proportionately high labor expenses because the vehicles are smaller.  Ridership has been low in several cities even by the standards of suburban coverage services. But there are reasons to be optimistic Bellevue may do better. This region also boasts the nation’s largest vanshare program at King County Metro. The economics of local vanshare programs are healthy with high fare recovery rates. For riders, CommutePool could provide many of the advantages of vanshare, but with greater flexibility to travel at different times or to use the service only on some days.

Existing shuttle buses at many Bellevue employers could be combined to capture economies of scale with a collaborative program. That adds some complexity in coordinating services with demand at multiple locations, but the cost efficiencies will make CommutePool appealing for employers.

For fixed-route transit providers, there is little downside. Bellevue’s service seems to compete mostly with SOV trips and other private shuttle services, and does not draw funding away from fixed services in other areas. Nevertheless, it’s critical that CommutePool be able to overcome the higher driver costs of smaller vehicles by operating with autonomous vehicles as soon as possible.

34 Replies to “Bellevue prepares for autonomous vehicle transit”

  1. Good to see Bellevue investing in the future. Half-hourly suburban coverage routes winding around subdivisions on their way to a transit center will never attract much choice ridership, and aren’t realistic alternatives for most suburb-suburb commutes.

    A network of frequent rail and BRT service between the major suburban activity centers, with autonomous neighborhood shuttles at stations, along with shared e-bikes and scooters to provide convenient last mile connectivity, makes a lot more trips realistic without a private car. And solves the scaling problems inherent in the park and ride model.

    1. Agreed, this is hopefully a compelling last-mile technology that can feed HCT lines and replace low ridership local routes.

      1. Hmm, on closer reading, this isn’t a last-mile solution, but an alternative to carpooling.

      2. Yeah, this particular implementation is coming online before there is an HCT network to feed, but it’ll get the technology established.

  2. So is the intent to pick up riders in Renton/Kent/Auburn and drive them all the way to Bellevue/Kirkland, or to connect them to 167-405 buses and simply provide last mile connections on both ends?

      1. No plan to connect to HCT – arguably it’s premature as the rail/BRT won’t be available until 2023/2024 and Bellevue is trying to move faster than that.

        Would be interesting to explore whether it makes sense after 2024. It does add more transfer friction to the service – if you’re already on an bus to the freeway, why not have the bus continue in the HOT lane to Bellevue. In some ways, this seems to be evolving toward a small-vehicle version of the branching BRT service many of us would have preferred over the highway-only BRT Sound Transit is building.

      1. Microtransit is not cost-effective. The economics of it rely on the assumption that the vehicles will be driverless. Driverless technology is not ready, and won’t be for many years, regardless of what people in the industry seem to be thinking. They will blow through the grant money before driverless cars are available, and be stuck with a high-cost niche transportation system.

      2. This is not “microtransit”; it’s an extension of vanpools. All the promotion materials talk about access to Bellevue jobs, probably peak hours, not for miscellaneous all-day trips. So it’s not really a last-mile solution or transit-alternative solution per se, it’s to attract the untapped commuter swarm that isn’t already on transit. If some people switch from Metro/ST to this, the hope seems to be that it will be only a few people, and those who live furthest from a transit station.

  3. Autonomous vehicle solutions can take many forms. It’s great that Bellevue is considering them! I still can’t wonder if Bellevue should first focus on Link access first in general though. Is Bellevue ready for the onslaught of shared vehicle and drop-off/pick-up trips that are coming to their Link stations in 2023? It may also be more effective to locate the receiving Link stations for autonomous vehicles to stations other than in Downtown Bellevue.

    In today’s “instant” or time-sensitive environment, we need to realize that buses every 30 or 60 minutes is becoming as antiquated as horse-drawn carriages were 100 years ago. Infrequent services for short distances usually means a whopping 1 or 2 percent mode share — a share that the Ubers and Lyfts of the world have even surpassed for some trips.

    I expect Mercer Island and Redmond to begin to explore last mile access relatively soon — culminating in 2023. I think that Shoreline, Mountlake Terrace, Lynnwood, SeaTac, Kent, Des Moines and others won’t be far behind as Link stations get closer to their opening dates.

    The lack of Seattle to pilot similar efforts is noted here. There are many places in Seattle where frequent bus service is unproductive. It’s ironic given the low density nature of many outer areas of Seattle that are about 1/2 to 2 miles from a Link station either today or by 2021.

    Personally, I’d like to see Seattle lead a targeted effort to have station access studies for every existing and under construction Link station (beyond tiny bus stops and no turn-around lanes for today’s infrequent buses).

    1. Seattle has Cartogo, Reachnow, and a vast fleet of Uber/Lyft drivers, not to mention 1000s of rentable bikes and frequent bus service on nearly every arterial.

    2. Bellevue has an excellent transit master plan but it’s waiting for somebidy else to pay for it. There may be a countywide transit measure by 2023, otherwise there may just be revenue-neural restructure. Either of those would be a significant improvement in frequency on the main corridors

  4. At first glance it seems to duplicate 405 BRT and Link’s Bellevue segment. But in reality the bulk of people in Eastgate/Somerset, Renton, Kent, and Auburn live far away from from Renton TC, Kent Station, Factoria Blvd, etc, because the residential areas are in east Renton, east Kent, and east Auburn, and transit from there to the nearest station is slow and infrequent. As to vans being less efficient on 405 and 167 than buses are, the total travel on those highways is enough for both: buses for people who live close enough to the bus stations at both ends, vans for those who don’t. They are only targeting a subset of the population: AM/PM commuters from southeast King County to central Bellevue employers, so it’s not replacing the mass-transit network but simply complementing it. And it is an extension of vanpools which have long been seen as a good thing.

    I’m skeptical of autonomous vehicles over such a large and open-ended area, but this is the kind of thing that could be a pilot for that. As could a few shuttle routes across Bellevue, where the burden of precisely mapping the route environment would be much less.

    1. Think I remember Sci-Fi taking care of this. Freeway vehicle is giant flat truck, with comfortable lounges for long-distance. Running the Interstates 24-7-365 without stopping.

      After receiving a signal, this 100 mph monster will lower a ramp to the fast-moving pavement, and the car’s electronics will drive the car onto the speeding ramp into the giant streamlined garage, from where the passenger can go upstairs and watch the miles go by.

      However….will not work on shoulder lanes, or HOV’s that come to an end north of SR520. So could take awhile. Which, for everything automated on wheels, is a good thing.


  5. 1. Where are these vehicles going to be tested, and by whom certified?

    2. Where will they be adjusted and maintained? Under whose supervision?

    3. Will the voters of Bellevue, and the rest of their work area, have any say about these vehicles sharing streets with their own cars. And families?

    4. Everything electronic is prone to sudden total failure, with no warning whatever. Including usual mechanical distress sounds and vibrations, and smell of something burning. How’s this truth to be handled?

    5. Who is going to deal with crashes and other distress calls? And communicate casualty information from the car, no matter how badly injured they are, or frightened, unconscious or dead?

    6. Who is going to investigate accidents?

    7. What is top speed going to be, and why are we not seeing it?

    8. How much extremely arduous testing will these cars have had before service starts? Like for a year before system carries first passenger?

    9. Will any of this testing duplicate, real-time operating in most arduous weather and traffic conditions? Including with cars full of people- at first, given helmets and flame-suits? Test personnel should be executives of the company that owns them.

    10. Will there be public official One to manage and supervise operations at all times- the same as for the rest of public transit?

    11. Since this is completely untried technology, which we taxpayers ourselves pay, ride, insure, defend in court, and have to share roads with- why not give it five years in public hands before we sent it private? Because when anybody dies, any private company will walk out of court with a suspender-equipped barrel for a business suit.

    ‘Way too many unknowns, ‘way too many variables. Many completely undiscoverable, let alone undiscovered. The hard way. So here’s minimum for me before passenger any passenger gets into a moving one. Only skilled techs in regulation Formula One racing gear, and a large fire extinguisher.

    Tough, simple, and with a dodge-em car cushion all the way around. In other words, close model for actual passenger service. And just to be on the safe side: Before we even start testing, let’s wait ’til we get the accident report for Train 501.

    Mark Dublin

    1. I have WAY more faith in a totally autonomous driver than I do in any of those people behind the wheel. Computers make tragic mistakes, but they are rare. People make little mistakes way more often, and those little mistakes frequently cause mortal injury. We call them “accidents” because they are just tiny little mistakes.

      Similarly, give me one massive plane crash instead of a normal year’s highway deaths.

    2. It goes both ways. People can be tired, distracted, stressed, or inebriated. Robots can fail to recognize whether an object or situation that’s usually safe or dangerous, is the opposite in this case, which humans can recognize and predict instantly.

  6. Some very basic questions:

    How many vehicles to produce 1000-3000 riders?

    How was that forecast done?

    Has the city done an EIS? Or a traffic analysis to show how these vehicles will affect arterial traffic, highway LOS, and transit reliability?

    Has the city consulted WSDOT on a public expenditure that would add usage to state facilities, and co-mingle robotic cars with human-operated cars, trucks, and buses?

    Has there been risk or safety analysis of such co-mingling?

    Are these FTA funds? And if so, does this service constitute public transit?

    If it does, the city is basically proposing to operate a competing service within the boundaries of a Public Transit Benefit Authority, and a Regional Transit Authority, which in turn means the city might — might — need the permission of both KCM and ST to proceed.

    The FTA has invested heavily in both agencies, and both take great pains to demonstrate to FTA that their investments are complementary and mutually beneficial. If such investments would put prior federal investment at risk, then it’s not likely FTA would proceed with Bellevue on something like this.

  7. Ok, come on! I left that one lying there for somebody to tear my logic to death for me. Train 501 wasn’t automated! So could be fair to say that if we’d had Automatic Train Protection, new trainee engineer could’ve been texting the whole dictionary while the robotics got speed down from eighty to thirty.

    But my point was that the automation question isn’t about machinery vs. people. It’s about what human designers, manufacturers, and users tell the machine, and each other, and users, and themselves about its operation. Worst outcome is when they’re wrong and don’t know it. For as many reasons as there are rocks in the gravel.

    Why would anybody put the track from Lakewood south any faster than forty? Maybe because somebody put it in the project program, to glamorize the new line. Fast. Or maybe somebody’s stipulation said they just had to to rush into compliance with some rule.

    And why would the railroad, and the trainmen’s union, and anybody’s legal department leave in place a curve they all knew was dangerous- over the budget! And then give train controls to someone who’d never driven that stretch before, as part of a training session involving a lot more people than him?

    Human or robot, “autonomous” is a lie, because some human had to tell it every single thing it needed to do. And when. From word or keyboard, taught wrong, goes wrong.

    ATP or not, everybody on duty knowing, and communicating how to get the job right would’ve kept 501 on the rails ‘way past the time they changed the number.


    1. >> Human or robot, “autonomous” is a lie, because some human had to tell it every single thing it needed to do. And when. From word or keyboard, taught wrong, goes wrong.

      But once the robot is told what to do, it does it. We tell human drivers what to do, and then they take selfies behind the wheel. I’ll take a software bug any day over if my other option is trusting a careless human.

  8. Just out of curiosity: How will they deal with fare enforcement? Do you have to tap your phone to get on the van? What is stop someone else from riding as well? Or is the plan to just use the honor system, and if so, why not extend that approach to the entire system?

    1. It says the vans will stop at major buildings, but it doesn’t say if it is only limited to employees going to those buildings. So, are customers or vendors or visitors also going to those buildings also allowed?

    2. If the computer is smart enough to drive itself, it ought to be smart enough to count the number of faces inside the van to detect cheats. The van doesn’t move when there’s an unpaid person inside. If someone is trying to cheat, social pressure from the other passengers will force that person to either pay up or get out.

      1. Any thoughts on what the law says about that, asdf2? Let alone, without legal authority, “forcing” anybody to do anything?

        And might someone also be able to create enough social pressure of their own to make the other passengers be the ones to leave if they know what’s good for them?

        Consequence of being too cheap to hire a driver. You might to have to pay somebody else to be aboard to collect or inspect fares to keep order. Like another person I definitely would never share a cab with- which might have to be autonomous because nobody can stand to get that close to him would put it:

        “Sad. Really sad. And dishonest, too.”


  9. It wouldn’t surprise me if they implemented a downtown circular in using this tech. I remember talking to someone from the City of Bellevue and they said that there has been talk of implementing one but it was too cost prohibitive to implement with traditional buses but said autonomous vehicles would probably address most of the issues.

  10. Easiest fare collection and enforcement would be at proof of payment shelters or building entrances, and/or enter through gates. Would limit number of fare-checking personnel.

    I don’t think “hailing” an autonomous cab on the street would be a very good idea for other reasons. Like a not getting a recorded view of who’s getting on, and what they’re about to do.

    But, you know, I’m a long way from conceding that this technology is anything I’ll tolerate on any square foot of pavement I intend to drive, bike, skate-board, or walk across.

    Getting same whiff of stale hustle-scented carnival popcorn as with delivery and surveillance drones. Let me vote, and put them in lanes with high enough barriers alongside to keep them from going anywhere else autonomously and I’ll think about it.


    Read it and puke while you weep. No, it’s not proof that automated is safer than driven. Just that whoever laid out this test never drove a golf-cart on a course where anybody else was playing golf.

    However fast someone’s reflexes while driving, no human can switch on vision from the automatic “off” setting that human consciousness goes into when it subconsciously thinks split-second attention isn’t likely to be needed.

    There’s no such thing as an “Emergency” human driver, who can engage and take control when a digital system fails with its usual lighting speed.

    While also knowing internally, and occasionally wrongly, that during average trip the human doesn’t expect to do anything but hold the wheel and relax. Meaning there’s one thing more dangerous than a robot driver.

    A human driver with absolutely nothing to do except in the last split second of somebody’s life, either their targets. their passengers’ or their own.



  11. Interesting application to help reduce the impact of too much traffic and stimulate the use of autonomous vehicles in the city.

Comments are closed.