As Google expands its self-driving car project into new US cities, its cars are coming across an increasing amount of obstacles – and not all of them avoidable.

The company is expanding its tests of autonomous vehicles outside the streets of Mountain View, California, for the first time – with Austin, Texas now home to one of its Lexus RX 450h cars.

The Lexus has been quietly driving around Austin for a couple of weeks, according to local news reports, but Google has now made its placement in Texas permanent.

Google has two types of self-driving cars – the Lexus, of which there are 23 on public streets, and a dome-shaped prototype car, of which two out of 25 have so far been unleashed on the public as of last month.

The prototype cars are speed-limited to 25 miles an hour (40km/h).

Google has been scaling up numbers of these prototype vehicles fast – from nine units in May 2015 to 25 in June, however most are being put through their paces on a private test track rather than public roads.

As more Google cars are put onto public streets, accidents will happen. As of May, Google has been reporting these scrapes and the reasons behind them, and in all cases it is quick to point out that “not once was the self-driving car the cause of the accident”.

The latest accidents both involved Lexus vehicles being rear-ended by other careless – human – drivers. It’s this kind of minor accident – as well as more serious ones – that a world of self-driving cars could cut substantially.

Prior, Google said its cars had been in 12 "minor" accidents from May 2010 to May 2015: again most were rear or side impacts caused by other drivers.

In one case, a Google car was the cause, but it was being driven manually, with the self-driving technology switched off.

Chasing ducks

Speaking at a TED conference earlier this year, Google’s self-driving car project director Chris Urmson said the company’s self-driving cars view the world around them geometrically.

Other cars, cyclists, pedestrians and obstacles are represented by different coloured shapes.

A Google car pauses on a green light while an emergency vehicle approaches.

“Once we started driving on neighbourhood and city streets, the problem [of self-driving navigation] becomes a whole new level of difficulty,” Urmson said.

“You see pedestrians crossing in front of us, cars crossing in front of us, going every which way, the traffic lights, crosswalks. It's an incredibly complicated problem”.

Urmson showed Google’s cars having to navigate roadworks where detours were marked with traffic cones.

He also said the cars had to recognise and react to the expectations of other road users.

“So, when a cyclist puts up their arm, it means they're expecting the car to yield to them and make room for them to make a lane change,” he said.

“And when a police officer stood in the road, our vehicle should understand that this means stop, and when they signal to go, we should continue.”

Google is solving these kinds of challenges with networking and data collection. For example, by sharing data between vehicles, cars that approach the traffic cones of the same road work later can automatically shift lanes.

However, engineers are still occasionally surprised by the types of obstacles cars have to deal with on public roads.

“Just a couple of months ago, our vehicles were driving through Mountain View, and this is what we encountered - a woman in an electric wheelchair chasing a duck in circles on the road,” Urmson said.

“Our vehicles were able to encounter that, slow down, and drive safely.”

Similarly, the car had to recognise and react to birds that flew across its path; to people who suddenly opened car doors and exited their vehicles without checking for approaching traffic; to emergency vehicles; and to cars and cyclists “blowing” through red lights.