Tesla finally blogged its pending release of its Autopilot Version 8 upgrade over the weekend with a detailed technical description of how radar will be used to avoid crashes.
CEO Elon Musk not long ago tweeted that Version 8 would be released by the end of Labor Day weekend. The Falcon rocket explosion and chronic lateness may have played into the delay.
Musk did tell reporters on Sunday during a conference call that the upgrade likely would have prevented the May fatality in Florida attributed to failure in the Autopilot system.
Tesla released details in its blog on Sunday describing how the Autopilot upgrade will be able to tap into its radar, rather than cameras, which will improve accuracy and rapidly respond to hazards. The key element is using more advanced signal processing from the onboard radar.
The radar was added to Tesla vehicles in October 2014, but only as a supplementary sensor to the primary camera and image processing system, Tesla said. After careful consideration, the company now thinks it can be used as the primary control sensor without relying on the camera to conform visual image recognition.
“This is a non-trivial and counter-intuitive problem, because of how strange the world looks in radar,” Tesla said.
The non-trivial and counter-intuitive part will be overcoming the limitations of radar, the company said. Problems coming from radar can include metallic objects functioning as a mirror; or humans appearing partially translucent; or something made of wood or painted plastic being as transparent as glass to radar.
One major hazard on roads would be a metal surface with a dish shape. It’s not only reflective, but it amplifies the reflected signal to many times its actual size.
“A discarded soda can on the road, with its concave bottom facing towards you can appear to be a large and dangerous obstacle, but you would definitely not want to slam on the brakes to avoid it,” Tesla said.
The biggest problem using radar is how to avoid these false alarms. Slamming on brakes to avoid a false hazard like a soda can creates unnecessary stress for the driver and, at worst, can cause injuries.
Software 8.0 now provides a three-part solution to these safety hazards, the electric automaker said.
Part one is have a more detailed point cloud providing more information on the objects.
“Software 8.0 unlocks access to six times as many radar objects with the same hardware with a lot more information per object,” according to Tesla.
Part two basically consists of creating a rapid-fire radar image slideshow. Radar snapshots take place every tenth of a second, creating a 3-D “picture” of the world, Tesla said. This will help the car discern whether an object is moving or stationary, or to distinguish inaccurate reflections.
“By comparing several contiguous frames against vehicle velocity and expected path, the car can tell if something is real and assess the probability of collision,” according to the blog post.
Part three takes on the tough challenge of making sure navigation data and height accuracy are taking an accurate reading and the Autopilot system doesn’t overreact. A rise in the road or driving under a bridge can throw everything off by looking like a collision course; it could also mean responding too late to brake when necessary.
Tesla said that “fleet learning” will come in handy at this point. The car can tap into the Tesla database to gauge positioning of road signs, bridges, and other stationary objects.
“The car computer will then silently compare when it would have braked to the driver action and upload that to the Tesla database. If several cars drive safely past a given radar object, whether Autopilot is turned on or off, then that object is added to the geocoded whitelist,” Tesla said.
The Tesla car can use that collective data to determine when mild braking should be used instead of unnecessarily slamming on the brakes. Braking force will gradually increase to full strength when it is “99.99% certain of a collision.” The crash may not be avoided entirely, but the impact speed will be dramatically reduced to point that serious injuries to vehicle occupants would be unlikely, Tesla said.
“The net effect of this, combined with the fact that radar sees through most visual obscuration, is that the car should almost always hit the brakes correctly even if a UFO were to land on the freeway in zero visibility conditions,” Tesla said.
Software 8.0 will protect the Tesla car occupants even if the driver of the vehicle in front can’t avoid the UFO.
“Taking this one step further, a Tesla will also be able to bounce the radar signal under a vehicle in front – using the radar pulse signature and photon time of flight to distinguish the signal – and still brake even when trailing a car that is opaque to both vision and radar. The car in front might hit the UFO in dense fog, but the Tesla will not,” Tesla said.
The blog post ends with additional release notes for those interested in the technical details, including approximately “200 small enhancements that aren’t worth a bullet point.”