Japan To Develop 3D Maps For Self-Driving Cars (nikkei.com) 25
An anonymous reader quotes a report from Nikkei: A joint venture in Japan will begin creating high-definition 3D maps for self-driving cars in September as part of a government effort to have such vehicles on the road by 2020, when the Tokyo Summer Olympics will be held. Tokyo-based Dynamic Map Planning, set up by Mitsubishi Electric, mapmaker Zenrin and nine automakers, will digitally chart the country's key expressways by driving a vehicle loaded with special surveying equipment. The data will be processed using computers designed for the creation of maps, which will be provided to automakers that invest in the startup. As a first step, Dynamic Map Planning will make maps covering 300km of the country's main expressways. The combination of high-resolution 3D maps and sensors will enable the accurate detection of which lane a car is in and the distance to junctions. High-precision surveying technology is required to make the maps, so Mitsubishi Electric developed equipment that will be installed on a canvassing vehicle. GPS will track the location of the car on the map, and sensors designed to detect the inclination of the car will measure the road grades. At the same time, data including the locations of road signs and traffic lights, as well as right- and left-turns and pedestrian crossings, will be collected using lasers. The survey data will be displayed as a collection of dots. Lines on the road, such as lanes, noise barriers and road signage, will be plotted on that image to faithfully re-create road conditions for 3D maps.
You all know what 3D maps mean.... (Score:2)
Flying cars, at long last!
Cut it? (Score:1)
lucky if it's just an subscription and not new car (Score:2)
lucky if it's just an subscription and your car can't run maps V 3.40 buy a new car!
Re: (Score:1)
Interesting times...
Is this more than a point cloud? (Score:2)
It sounds like it is just a laser scan with a point cloud, albeit on a very substantial scale. Am I missing something?
I do wonder how they factor out the other cars though...
Re: (Score:2)
This will be more than a LIDAR point cloud. IMU data/ RTK GPS/ Images and photogrammetry all need to be included. LIDAR relies on images to collect color data, and those images will be used to reference lines in the road etc. Very little data can be extracted directly from the point cloud without making the a polygonal mesh.
high resolution today (Score:2)
3D maps can't be the only source (Score:4, Insightful)
Re: (Score:2)
My car map data is about 4 years old now. There are times it thinks I am driving in fields where roads got shifted. Kind of funny. A true test of selfdrive would be during road construction. In texas both 35 and mopac in austin stress even a human driver. There are marks from partial lane repave, old lane markers, new lane markers and newer lane markers. There are times I have trouble determining which mark I should follow. I would not want to be near an auto-piloted car in these sections.
Re: (Score:2)
having near survey grade maps is not a the end of the navigation assistance, it is a start however. Autonomous cars wil have rtk gps, lidar, radar and computer vision all working in tandem with highly accurate 3d maps to do situational avoidance.
Re: (Score:2)
autonomous cars will collect data as they drive. All autonomous cars are capable of mapping to a degree.
Re: (Score:2)
No one is suggesting that autonomous cars are wholly reliant on 3d maps they have in their database. They will have sensors to read street signs, and new road shifts etc. The 3d maps assist in the process but are not the only means for guiding the vehicle. The sensors on autonomous vehicles can easily be used to update changes in the map. Fleets of vehicles will be able to upload to servers and using the magic of modern tech they will be able to be updates in the vehicle.
Re: (Score:2)
No one is suggesting that autonomous cars are wholly reliant on 3d maps they have in their database. They will have sensors to read street signs, and new road shifts etc. The 3d maps assist in the process but are not the only means for guiding the vehicle. The sensors on autonomous vehicles can easily be used to update changes in the map. Fleets of vehicles will be able to upload to servers and using the magic of modern tech they will be able to be updates in the vehicle.
So why the need for this special map making tour? It's because they don't plan on using SLAM or fusing it into a global database this decade and probably not the next. They are implementing lane following and some minimal traffic rule following which they cannot automate or there wouldn't be a need for this manual process.
Re: (Score:2)
Locations such as traffic lights are derived from the map, not real time slam. This allows the vehicle to know where to look to get the appropriate information without having to scan the area each and every traffic light.
I saw a rig like this in 2005 (Score:2)
When I was shooting Fast and Furious: Tokyo Drift in Tokyo, in winter of 2005, I saw a rig a lot like this. I was walking down a street, and saw a van with four cameras, four LIDARs, and two GPS sensors on the top. I asked a man working on it what it was, and he said "Oh! Are you engineer?" and I confessed that I was just a movie-maker. But, nevertheless, they showed me everything in the van, and said that the point was developing 3D models of all the streets in Tokyo. At the time, it wasn't for self-d
Why 3D? (Score:2)
Why do they need 3D maps to determine "what lane they are in and distance to the next junction"?
Re: (Score:3)
Re: (Score:2)
How should the vehicle know how many lanes are on the road without that data being made available in a map? You can't have a frame of reference and say I'm in the third lane if you don't know how many lanes there are. Also vehicles etc will block the ability for a lidar scan to see how many lanes are on the other side of that truck.
Inefficient (Score:2)
Seems a rather inefficient way to track it. Wouldn't it be more efficient to do the mapping from plane?