selfdrivingcar   245

« earlier    

Uber insiders describe infighting and questionable decisions before its self-driving car killed a pedestrian
This is really bad in several ways. 1) there's a reason why you have "pilot flying" and "pilot not flying" or "pilot monitoring" even when the AP is on. 2) making nonintervention a goal penalises intervention 3) making nonintervention a goal also rewards inattention and also machine insensitivity until it hits somebody 4) sterile cockpit - if you're treating it as relaxing time this is going to reinforce leaving it as late as possible to intervene 5) sure it's software, but the hardware team also seems to be the car team and it's a car 6) it's not ready to be on the road and it shouldn't be on the road
selfdrivingcar  safety  autopilot  uber  to_blog 
12 days ago by yorksranter
felipessalvatore/self_driving_pi_car: A deep neural network based self-driving car, that combines Lego Mindstorms NXT with the computational power of a Raspberry Pi 3.
A deep neural network based self-driving car, that combines Lego Mindstorms NXT with the computational power of a Raspberry Pi 3. - felipessalvatore/self_driving_pi_car
selfdrivingcar  deeplearning  raspberrypi 
20 days ago by ElliotPsyIT
Navigate on Autopilot is Useless (2018.42.3) | Tesla Motors Club
The car regularly suggested lane changes directly into objects it clearly detected. It would even show the proposed path on the visualization as going directly through the other vehicle. In one instance I wondered if it really was going to let me change lanes into a semi truck, or if it would wait until it was clear. Nope, it started to move right towards it after confirmation. No red lane, nothing, while directly along side a semi.
tesla  selfdrivingcar  bubble2 
25 days ago by yorksranter
Volkswagen a winner as EU set to favour wifi over 5G: draft
EU version of DSRC vs C-V2X debate. ECommission doesn't want C-V2X. Telcos gonna whine.
5g  dsrc  selfdrivingcar  europe  ovum 
6 weeks ago by yorksranter
Vigilante engineer stops Waymo from patenting key lidar technology • Ars Technica
Mark Harris:
<p>Following a surprise left-field complaint by Eric Swildens, the US Patent and Trademark Office (USPTO) has rejected all but three of 56 claims in Waymo's 936 patent, named for the last three digits of its serial number. The USPTO found that some claims replicated technology described in an earlier patent from lidar vendor Velodyne, while another claim was simply "impossible" and "magic."

Swildens, who receives no money or personal advantage from the decision, told Ars that he was delighted at the news. "The patent shouldn't have been filed in the first place," he said. "It's a very well written patent. However, my personal belief is that the thing that they say they invented, they didn't invent."

The 936 patent played a key role in last year's epic intellectual property lawsuit with Uber. In December 2016, a Waymo engineer was inadvertently copied on an email from one of its suppliers to Uber, showing a lidar circuit design that looked almost identical to one shown in the 936 patent…

…Remarkably, Swildens does not work for Uber or for Velodyne, nor for any other self-driving developer—he works for a small cloud computing startup. Swildens became interested in the patent when it surfaced during the Uber case, and he saw how simple Waymo's lidar circuit seemed to be. "I couldn't imagine the circuit didn't exist prior to this patent," he told Wired last year.

Swildens' research uncovered several patents and books that seemed to pre-date the Waymo patent. He then spent $6,000 of his own money to launch a formal challenge to 936. Waymo fought back, making dozens of filings, bringing expert witnesses to bear, and attempting to re-write several of the patent's claims and diagrams to safeguard its survival.

The USPTO was not impressed. In March, an examiner noted that a re-drawn diagram of Waymo's lidar firing circuit showed current passing along a wire between the circuit and the ground in two directions—something generally deemed impossible.</p>

As everyone on Twitter has been saying, not all heroes wear capes.
selfdrivingcar  patent  lidar 
9 weeks ago by charlesarthur
Fully driverless Waymo taxis are due out this year, alarming critics • Ars Technica
Timothy Lee:
<p>Waymo, Google's self-driving car project, is planning to launch a driverless taxi service in the Phoenix area in the next three months. It won't be a pilot project or a publicity stunt, either. Waymo is planning to launch a public, commercial service—without anyone in the driver's seat.

And to date, Waymo's technology has gotten remarkably little oversight from government officials in either Phoenix or Washington, DC.

If a company wants to sell a new airplane or medical device, it must undergo an extensive process to prove to federal regulators that it's safe. Currently, there's no comparable requirement for self-driving cars. Federal and state laws allow Waymo to introduce fully self-driving cars onto public streets in Arizona without any formal approval process.

That's not an oversight. It represents a bipartisan consensus in Washington that strict regulation of self-driving cars would do more harm than good.

"If you think about what would be required for some government body to examine the design of a self-driving vehicle and decide if it's safe, that's a very difficult task," says Ed Felten, a Princeton computer scientist who advised the Obama White House on technology issues.</p>

Pretty much impossible to prove "safe". But how safe? Safer than a human? My suspicion is that they will be safer than humans in general, but do some strange things leading to accidents when humans wouldn't have.
regulation  waymo  selfdrivingcar 
10 weeks ago by charlesarthur
This military tech could finally help self-driving cars master snow • Ars Technica
Jonathan Gitlin:
<p>The research conducted at the country's National Laboratories is usually highly classified and specifically aimed at solving national security problems. But sometimes you get a swords-into-ploughshares moment. That's the case here, as a startup called WaveSense looks to apply technology originally developed by MIT Lincoln Laboratory to detect buried mines and improvised explosive devices for use in self-driving cars.

If you want a car to drive itself, it has to know where it is in the world to a pretty high degree of accuracy. Until now, just about every variation of autonomous vehicle we've come across has done that through a combination of highly accurate GPS, an HD map, and some kind of sensor to detect the environment around it. Actually, you want more than one kind of sensor, because redundancy is going to be critical if humans are going to trust their lives to robot vehicles.

Most often, those sensors are a mix of optical cameras and lidar, both of which have pluses and minuses. But is a combination of lidar and camera truly redundant, if both are relying on reflected light? Other solutions have included far infrared, which works by detecting emitted light, but WaveSense's approach is truly photon-independent. What's more, it's the first sensor we've come across that should be almost completely unfazed by snow.

That's because it uses ground-penetrating radar (GPR), mounted underneath the vehicle, to sense the road beneath—now you can see where the military application was. The GPR scans the ground underneath it to a depth of around 10 feet (3m), running at a little over 120Hz to build up a picture of the subterranean world beneath it. As the car drives along, it compares that data to a map layer of already-collected GPR data for the road network and can place the car to within a couple of centimeters.

Yes, this requires pre-mapping, but so does lidar. And WaveSense says that remapping should be far less frequent as conditions under the road are less subject to change than they are above ground.</p>

OK, but don't we need them to master fair-weather roads first?
selfdrivingcar  snow 
august 2018 by charlesarthur
Hackers plan to keep GM's self-driving cars safe
Their design calls for the car to refuse any inbound connections — no data will come to the vehicle unless it asks for it first // much dozy selfdrivingsoylent bollocks, but good point on that one.
selfdrivingcar  security  ovum 
august 2018 by yorksranter
Hackers plan to keep GM's self-driving cars safe • Yahoo News
Rob Pegoraro:
<p>Their plan for the autonomous vehicles coming from Cruise, based on the Chevy Bolt electric car, starts with a simple premise: Remove the systems that opened up those other vehicles to remote attacks.

Bluetooth? Forget it — the car is driving itself, so you don’t need hands-free calling. The radio? You’ll listen to your phone anyway. And that fancy touchscreen hardwired into the dashboard doesn’t need to exist either, not when the passengers can interact with the car via a stripped-down, locked-down tablet.

“If you don’t need something, take it out,” Valasek said. It’s Security 101 to reduce a device’s “attack surface” — the parts that respond to outside inputs, and which an adversary could therefore try to exploit. But it hasn’t always been Connected Car 101.

Miller’s and Valasek’s formula also includes a healthy dose of paranoia. Their design calls for the car to refuse any inbound connections — no data will come to the vehicle unless it asks for it first.

And much as in the locked-down framework Apple (AAPL) built for the iOS software inside iPhones and iPads, this autonomous-vehicle system will digitally sign and verify code at all levels, with messages from one component to another encrypted whenever possible.

Miller noted one possible speed bump: The wired networking in many cars is too old to support that encryption. “The components in cars are just so far behind,” he complained.

If this level of security by design sounds like something worth paying extra for — sorry, you can’t. Cruise Automation will run only as a ride-hailing service, like an Uber or Lyft but devoid of life forms in the driver’s seat.

That solves the issue of how you sell a car without a radio or Bluetooth: You don’t have to.</p>

Clever - and probably necessary.
selfdrivingcar  hacking 
august 2018 by charlesarthur
The dream of driverless cars is dying
there were few cars and certainly none that were remotely driverless. One, a huge SUV, was there solely to demonstrate equipment that enables a driver to point at a building to find out its purpose, and if it were a restaurant to book a table, and if a cinema, to buy a ticket. Lars, who was demonstrating this with great enthusiasm, tried to convince me it would change my life.
july 2018 by yorksranter
Former Apple employee charged with theft of trade secrets related to autonomous car project • Mac Rumors
Juli Clover:
<p>Xiaolang Zhang was hired at Apple in December of 2015 to work on Project Titan, developing software and hardware for use in autonomous vehicles. Zhang specifically worked on Apple's Compute Team, designing and testing circuit boards to analyze sensor data.

He was provided with "broad access to secure and confidential internal databases" due to his position, which contained trade secrets and intellectual property for the autonomous driving project that he ultimately ended up stealing.

In April 2018, Zhang took family leave from Apple following the birth of his child, and during that time, he visited China. Shortly after, he told his supervisor at Apple he was leaving the company and moving to China to work for XMotors, a Chinese startup that also focuses on autonomous vehicle technology.

Zhang's supervisor felt that he had "been evasive" during the meeting, which led Apple's New Product Security Team to begin an investigation, looking into Zhang's historical network activity and analyzing his Apple devices, which were seized when he resigned.

Apple found that just prior to Zhang's departure, his network activity had "increased exponentially" compared to the prior two years he had worked at Apple. He accessed content that included prototypes and prototype requirements, which the court documents specify as power requirements, low voltage requirements, battery system, and drivetrain suspension mounts.</p>

Arrested at the airport as he was about to leave for China. Neil Cybart has <a href="">dug into the court filing</a>, which shows there are 5,000 Apple employees who know about "Project Titan" (the self-driving vehicle project) and 2,700 who have access to the Project Titan database. Here's the <a href="">full court filing</a>.
apple  selfdrivingcar 
july 2018 by charlesarthur
The dream of driverless cars is dying • The Spectator
Christian Wolmar went to a giant "Self-driving vehicle" exhibition in Germany, but found them in short supply:
<p>Surprisingly, I met more doomsayers than purveyors of the autonomous driving dream. The starkest warning came from Tim Mackey, who styles himself ‘senior technical evangelist’ for Black Duck Software, a company that specialises in security issues around autonomous vehicles. He believes there will be a seminal event that will stop all the players in the industry in their tracks. ‘We have had it in other areas of computing, such as the big data hacks and security lapses,’ he said, ‘and it will happen in relation to autonomous cars. At the moment, none of the big players are thinking properly about security aspects and then they will be forced to.’ He pointed to a video showing on another stand in which a man was calling up a car from a garage using a phone app: ‘That sort of thing is just too easy to hack. There’s all sorts of software put into cars that is old and easy to access. We just have to hope that the wake-up call will be minor and not kill anyone.’ Indeed, in a test a few years ago, hackers were able to get hold of a car’s steering and braking systems and Mackey is convinced that criminals will one day use the same method.

More widely, there was a general expectation these suppliers were riding the crest of a wave that will hit the rocks soon. While there is no doubting the scale of this industry, with billions being invested every year, none of the OEMs has yet made a penny from selling a driverless car. This money, benefiting these exhibitors, is therefore a punt, a high-stakes bet there is a pot of gold at the end of the rainbow. One, Johannes, told me: ‘I see a pattern like the dotcom boom. At some point, people are going to realise that the day when they start to get returns for their investment is far off, if ever. Then they will start pulling out and who knows how bad it will get. But the clever money will move somewhere else.’ The bad publicity caused by a couple of deaths in Tesla cars while its autopilot was engaged and by the Uber fatality may be seen as the start of public disenchantment with the concept.</p>

The Spectator is a fairly right-wing magazine, so you might expect it to be down on new tech; but I worked with Wolmar at The Independent, and he's fair but firm on topics like this.
july 2018 by charlesarthur
63. Russian Ground Battlefield Robots: A Candid Evaluation and Ways Forward – Mad Scientist Laboratory
• Instead of its intended range of several kilometers, the Uran-9 could only be operated at distance of “300-500 meters among low-rise buildings,” wiping out up to nine-tenths of its total operational range.

• There were “17 cases of short-term (up to one minute) and two cases of long-term (up to 1.5 hours) loss of Uran-9 control” recorded, which rendered this UGV practically useless on the battlefield.
russia  tank  robot  selfdrivingcar  ew 
june 2018 by yorksranter
Uber test car driver streamed Hulu before fatal crash • Consumer Reports
Jeff Plungis and Keith Barry:
<p>The Tempe police report says distraction was a factor in the crash that killed the pedestrian, Elaine Herzberg.

During Vasquez’s ride in the Uber vehicle, which was recorded on video inside the vehicle as part of the testing, she looked down 204 times, mostly in the direction of the lower center console near her right knee, according to the police report. She was looking down for 5.2 of the final 5.7 seconds prior to the crash, the report says.

A log of Vasquez’s account provided by the video-streaming service Hulu, under a search warrant, showed that “The Voice” was streaming on her account in the final 43 minutes of the drive and that the streaming ended at 9:59 p.m., the approximate time of the collision, the police report says. 

The police concluded that the crash wouldn’t have occurred if Vasquez had been paying attention to the roadway, and indicated that she could be charged with vehicular manslaughter. Details from the police report were published Thursday by the Arizona Republic, Reuters, and other media outlets.</p>

In which case what's the point of it being "self-driving"? It's the limitations that make this pointless. You couldn't trust it on motorways, side roads, at night. In which case there's no point having it. Self-driving systems have to be really, really good, or else not employed at all, because driver inattention will be a thing, and accidents will keep happening.
selfdrivingcar  uber 
june 2018 by charlesarthur

« earlier    

related tags

5g  africanamericans  ai  apple  ar  automation  automobile  automobiles  automotive  autonomous  autonomouscars  autonomousdriving  autonomousvehicles  autonomy  autopilot  benedictevans  bigdata  bubble2  busdrivers  car  cars  cities  companies  competitions  cupertino  data  datascience  datasets  deeplearning  diyelectronics  drive  driverless  driverlesscars  dsrc  dyson  elonmusk  ethics  europe  ew  google  gta  hacking  hardware  humancontrol  ia  illusion  imagesegmentation  intel  iot  jamesnozar  kaggle  keras  killing  lasvegas  lawsuit  lidar  lists  lyft  machinelearning  ml  mobility  openpilot  ovum  paloalto  patent  pedestrian  planning  pledge  privacy  python  raspberrypi  reality  regulation  robot  russia  safety  security  selfdriving  selfdrivingcars  selfdrivingsoylent  sillyvalley  snow  software  tampa  tank  tech  technology  tempe  tesla  to_blog  traffic  trucks  uber  udacity  use  video  waterstreettampa  waymo 

Copy this bookmark: