Drive.ai — a startup developing an on-demand autonomous shuttle service — is based in Silicon Valley but has deployed its first vans in Texas, drawn in by the state’s favorable regulations.

Without a national regulatory framework for autonomous vehicles, states have become laboratories not just for the technology itself but also for the rules emerging to shape it, Axios' Kaveh Waddell and Kia Kokalitcheva report.

• Their efforts will decide which Americans get to hitch a ride in an autonomous car first — and which cities will reap the potential economic benefits of self-driving vehicles, along with any problems.

The big picture: Federal legislation that would create national parameters for testing and deploying AVs passed the House but is stalled in the Senate, leaving states to create their own rules for now. Automakers worry that without federal standards they'll have to deal with a patchwork of state laws that would hamper a broader roll-out of the technology.

That leaves companies shopping for testbeds among the states. Currently, 29 states have passed some kind of regulation for self-driving cars, some more extensive than others.

• For Drive.ai, Texas strikes a Goldilocks medium between highly regulated states like California and those with lax rules, like Arizona, CEO Sameep Tandon says.
• Waymo, Uber and GM Cruise are testing in states like California and Pennsylvania where they're located and can experience urban environments.
• Others are operating in Arizona and Florida, where lighter regulations make it easier to test and deploy ride services.
• Testing is also determined by weather, not just regulations, National Conference of State Legislature’s Douglas Shinkle points out. So far, much of it takes place in warmer, sunnier climates like California, Nevada, Texas and Florida, where conditions are easier for today's AVs to navigate.

As companies get closer to deploying self-driving cars that can be available to the public, state rules around consumer transportation will be even more important.

• “Commercial deployment is the next big battleground in AV regulations,” says Greg Rogers of Securing America’s Future Energy.

Yes, but: Some experts argue that self-driving safety regulation should really be done at the federal level.

• “You should be able to buy a car in California and drive it to New York,” Rogers says, adding that the National Highway Traffic Safety Administration is best equipped to do this.

Go deeper: Read the entire story.

Human drivers are faulted for most accidents involving AVs, but closer inspection reveals the responsibility for such incidents is more complex and shared, National Auto Care's Steve Verney writes for Axios.

Many accidents appear to have occurred when an AV took an action that was technically legal and safe, but which a human driver did not expect.

The way AVs drive, at least for now, is fundamentally different from the way humans do, which may create inefficient and dangerous miscommunications.

• Machines learn by rules, algorithms, data-driven decisions and visual inputs of object and trajectory — all guided by the prime directive of avoiding objects in their path.
• Humans rely on different principles and experiences, as well as subtler cues from the environment, drivers and pedestrians (for example, the head nod that signals the go-ahead to merge).
• Humans also operate in a risk-reward setting — balancing the odds of an accident, say, with the odds of missing a flight.

What's needed: Public policy should ensure that autonomous development includes ways for humans to better identify and interact with AVs.

• While a lidar unit on top of an AV is usually a dead giveaway, a Tesla on autopilot is not.
• If AVs can communicate their intentions and next moves, humans will better respond.

The bottom line: AV developers, manufacturers, insurers and regulators will have to devise a more specific, thought-out vision for how humans and machines share the road. If they don't, public resistance could well impede the AV transition.

A recent study from the American Automobile Association found that 73% of Americans said they would be too afraid to ride in a driverless vehicle and 63% would feel unsafe sharing the road with one, May Mobility's Alisyn Malek writes for Axios.

Why it matters: Mistrust of robotic decision-making, and doubts about safety, are changing the focus of the AV industry. Increasingly, the critical question for companies is shifting from who can build the technology first to who can win people’s trust.

What to watch: Gaining the trust of riders, transit planners and government regulators will require transparency, education and in-person interactions with the technology. 

• These opportunities range from single-day demonstrations to pilot program deployments — partnerships between AV developers and communities that can help to educate the public, as the cities of Las Vegas and Columbus are already doing.
• Given the potential upsides of AV technology, the industry should not overlook the impact that early missteps could have on its development. With incidents like those in California and Arizona, many in the public are rightfully questioning if the technology is, or will ever be, ready. 
• In addition to communicating their own safety standards and performance, companies can also remind the public of similarities between AV technology and unmanned robotic systems that operate safely in everything from elevators to airport trams.

The bottom line: To win public trust, AV developers will have to take up these exposure and education challenges alongside the technology itself.

• The race for the next billion cars (Steve LeVine — Axios)
• Germany launches world's first autonomous tram in Potsdam (Kate Connolly — The Guardian)
• When will AVs be safe enough? An interview with Missy Cummings (Jeremy Martin — Union of Concerned Scientists)
• German carmakers quietly test AVs in China (Wang Xueqiao and Tom Hancock — FT)
• A dedicated chip design for autonomous cars (Eric Auchard — Reuters)

PerceptIn, a startup working on vision systems for robots and self-driving cars, recently announced it will offer a DIY custom autonomous vehicle, per Forbes' Jennifer Kite-Powell.

The price tag: $40K

How it works: The vehicle uses computer vision, radar, sonar, and GPS to sense its surroundings and plan a driving path. Customers can buy the vehicle parts to assemble by themselves or PerceptIn will design and build it for them.

The purpose: The pod isn't currently meant for the open road, but for controlled environments like university campuses where it can transport people and objects at low speeds (right now, the max is 20 miles per hour).

"We've created a Lego-like approach and designed the vehicle so that when people are putting it together, they can reconfigure different components to use only what suits their needs."

Yes, but: The vehicle's planning and control algorithm can get stuck when both lanes on the road are blocked, requiring a human to take control of the vehicle remotely.

What's next: Liu tells Axios they plan to apply the modular approach to other types of vehicles — including an autonomous vending vehicle to be announced soon.