A Ford Argo AI test vehicle drives through the downtown area in Detroit, Mich., last month. Such vehicles need to safely operate while facing poor road conditions, complex street networks and unpredictable human nature. (Jeff Kowalsky/AFP/Getty Images)

Autonomous car predictions are overly optimistic. Questions include how soon, how beneficial and to what degree American cities, urban land use, real estate development and transportation will be transformed. Lots of metaphorical rumble strips and roadblocks exist en route.

Countless millions of non-autonomous vehicles are on the road while tens of millions of new non-autonomous vehicles are manufactured, marketed, sold and leased annually. Moreover, round-the-clock access to and use of a personal vehicle, whether owned or leased, will always appeal to some Americans who like being in control, who like piloting a car loaded with features that make driving comfortable, safe and even fun.

Yet autonomous vehicles must be relatively ubiquitous because communicating with other autonomous vehicles is key to safe and effective autonomous operation. Consequently, phasing out non-autonomous vehicles, necessary for sufficient vehicular intercommunication, must be staged over time region by region. Thus, dominant autonomous car production and use throughout the country are years away.

Autonomous 18-wheelers already travel on multilane, city-to-city highways, although always with a driver on board for truck route starting point and endpoint navigation, and for coping with unforeseen problems. The generally straightforward, uninterrupted, geometry of major U.S. highways makes autonomous trucks feasible and, by reducing the risk of truck drivers falling asleep, safer.

But autonomous truck travel on highways is technologically easier to implement than having cars drive themselves. Off the interstates, vehicles must operate on local urban and suburban street networks that are much more diverse and complex than highways.

Parts of Boston, New York City, the District, Houston, Salt Lake City and Los Angeles encompass rationally gridded networks with readily navigable straight streets. But elsewhere in these cities are neighborhoods with winding roads, narrow streets and lanes, streets in bad condition, cul-de-sacs, blind turns and dead-ends. Topping the list are Boston’s historic downtown streets based on cow paths.

Riding autonomously through the street grids of Capitol Hill, Georgetown and Old Town Alexandria seems reasonable. But Arlington’s crazy, not-so-reasonable multistreet intersections are always challenging for the nonlocal human brain to comprehend. Will AI do better?

Equally challenging are traffic circles in Washington. Drivers must look continually in every conceivable direction, except maybe up, while anticipating — or guessing at — the second-by-second intentions and moves of other drivers, bikers, scooter riders and inattentive pedestrians focused on mobile phones. If later this century all vehicles and humans entering, rounding or crossing circles automatically communicate with one another, navigating circles may become simpler and safer.

An autonomous vehicle continuously scans, detects, processes and, microsecond by microsecond, responds to what it sees or senses under, around and above the vehicle. Even weather is factored in. Accordingly, the entire roadway environment must be physically in good condition and always visible.

Regrettably, this is frequently not the case. Lane markers and dividers, pedestrian crosswalks, stop lines and directional movement arrows can be faint or invisible and undetectable by sensors. Traffic signals malfunction periodically, and stop signs may be obscured by vegetation. Deteriorating, misaligned or missing curbs can confuse sensors when a car is looking for the road edge or is parallel parking.

With potholes showing up every winter, will autonomous vehicles see and miss them to avoid blowing a tire?

For many counties and municipalities, diligent monitoring and constant maintenance to avoid roadway deficiencies are costly and difficult to reliably ensure. This is why adverse roadway conditions will be among the roadblocks autonomous cars will face.

City form and function are predicted to change substantially because of autonomous vehicles. Reduced private car ownership and use, along with autonomous car sharing, could result in far fewer vehicles on the road. Assuming increased transit and other mobility options, the need for on-street parking, surface parking lots and parking garages is expected to diminish, thereby freeing up urban and suburban real estate for more desirable uses, including public parks.

But neither a city’s overall image nor the character of its public realm, defined in part by its streets, will be radically transformed. And vehicular activity will not disappear. No matter whether man or machine does the driving, multifunctional streets will continue to exist: for emergency and police vehicles; for trucks making deliveries; for building construction and maintenance service trucks; for moving vans; and for trucks collecting trash and recyclables. Bicycles, motorbikes and maybe scooters also are here to stay.

Short-term curbside parking to serve visitors and customers heading to sidewalk-level destinations will probably continue, as will automobile, limo, taxi and bus traffic. Millions of homes will continue to have garages. And public utilities will remain under, on or above street rights of way.

One other significant roadblock deserves mention.

Laws, regulations and standards governing roadways, vehicles and vehicle insurance today vary from state to state. This affects safety, makes interstate travel riskier and can entail burdensome, costly litigation. Feasibly manufacturing and marketing autonomous cars across the United States requires a coordinated, multi-jurisdictional effort to address governmental, institutional and legal inconsistencies.

My wife and I would love riding around Washington in an autonomous car, heedlessly reading or catching up on sleep. But this will be possible only in the distant future. Given the many potential roadblocks, it may not even be possible in our granddaughters’ futures.

Roger K. Lewis is a retired practicing architect, a University of Maryland professor emeritus of architecture and a guest commentator on WAMU’s “The Kojo Nnamdi Show.”