The future of the automotive field and autonomous field
Barak Rosenberg
I am an accomplished AI specialist, founder, entrepreneur, manager, programmer, architect. With a rich background in the hi tech industries, management, architecture, programming, AI/ML (including LLM edge/cloud).
The future of electric vehicles is known, the technology of electric vehicles is already proven, electric vehicles have very simple engine, much more effective which depends on a good battery. The efficiency of an Electric Vehicle is much better then any vehicle driven by a combustion engine.
This is not the case in the autonomous vehicle field and safety systems for driven vehicles, which is stuck, due to cost problems, complexity of the technology, and conservative thinking by vehicle manufacturers called OEM's.
Today, the safety of vehicles relies on sensors like radars/cameras, which cost a lot of many, do not recognize everything on road, and do not work in every weather, have problems with small objects far away, and can not recognize lanes especially in junctions where there are no lane markings on road.
The answer is not in using more sensors, or better GPU(costs very much, and consumes much electricity, is it really needed?), that uses same AI visual recognition software call NN(Neural Networks), which has it's problems, and has big problems with accuracy of what it recognizes, and much more important can not calculate exact distance to objects, even when the objects are recognized.
As we all use Google maps/Waze or other navigation software, a new map called HD-map which contains 50 times more location points of lanes then traditional map, with several more layers, including speed limit of every road must be used as a basis for recognizing lanes, and the exact speed limit in the exact place inside lane where vehicle is moving, so navigation of vehicle will use exact lane coordinates instead of depending on recognition of lanes which can be hidden by other vehicles, weather, or totally not defined on road.
The leading company in the world in the automotive field, will start to introduce a new technology in vehicles using movement on road and inside lanes based on HD-map and localization in 1-2 years, according to them.
The moment we have an HD-map with lane coordinates, and localization-knowing the longitude/latitude on earth where vehicle is moving(best localization we currently have is around 1meter using GNSS satellites. In the future 5G antennas and cellular communication will be used instead of GNSS satellites to reach accuracy of <10cm).
Sensors like cameras/radars will be used only for stopping the car in emergency, even when not being able to recognize the object ahead, still the car will be able to hold distance ahead using radars. Using an HD-map and localization with a radars/cameras will give a much better cruise control system then we have today that will be able to hold the car inside the lane including in junctions in every circumstances, and in every weather.
In order to move between lanes, a vehicle must know where it's going using global path planning, and once it knows the lane it is moving in, it can calculate it's local path planning between lanes and send the local path plan(50 points of vehicle ahead where vehicle will be in every 0.1 second in the future) to all cars around that have a V2V modem using wireless communication(V2V- Vehicle to Vehicle wireless communication using DSRC/C-V2X protocols).
Since there is going to be a mix of vehicles equipped with V2V modems and those without V2V modems. The vehicles that have V2V modem will have next level of cruise control using HD-map, and can move together into one lane to move together using Platoon V2V synchronization-as part of mixed environments.
Movement on road must be precise using HD-map lane coordinates with an accuracy of <10cm, using localization taken from 5G antennas(that will act as V2I devices) to have much better cruise control to hold vehicle speed and wheel in every circumstance, after that implementations of V2V communication will be used to get V2V enabled cars into one lane, and moving between lanes while broadcasting local path plan between lanes. All these developments will take several years...
To conclude: precise HD-maps must be developed with good localization(<10cm accuracy) to have a real cruise control system that is working in any weather and in any circumstances and turning inside junctions, in the next stages, adding V2V broadcasting to modems and local path plan ahead will give the ability to move between lanes.
The Vehicle must have best Situational Awareness of where it moves, where it is heading, including inside lane, and between lanes, and be able to broadcast local path plan/route to all cars around, also using the HD-map monitor system to display all vehicles around and test own movement and all other cars movement in real time.
See our HD-map&V2V monitor system integration : https://youtu.be/B0R5dClpl2g , this film that runs code includes displaying a real time HD-map on a monitor system with lane coordinates, location from GNSS satellites using V2V modem- to put vehicles inside the right HD-map tile->road/link->lane->lane segment, test speed limit of own vehicles and vehicles around on same HD-map tile, test movement of all cars around that broadcast BSM(Basic Safety Message that includes Longitude/Latitude/Speed/Heading) between all vehicles around equipped with V2V modems.
The initial step of putting V2Vmodems inside driven vehicles was done by NXP inside VW golf 8 cars, see: https://investors.nxp.com/news-releases/news-release-details/volkswagen-and-nxp-deliver-safety-european-roads-worlds-largest
This is just the first step in a long way that will lead us to a much better traffic system on roads, and to smart cities...
What we have today using ADAS sensors is merely ~5% of what is needed to have much better safety on roads, and help solve the traffic congestion problems in the world(Platoon synchronization of movement of several V2V enabled cars in same autonomous lane or 2+ passengers lane that we have today in many cities and countries)
This is the ONLY possible solution to handle traffic safety and traffic congestion problems that is in development today, and can work, it will take several years to develop...........
We are leading the field of traffic management field using HD-map(which is now in development and will be integrated in vehicles in the next 1-2 years..), with V2X communication into the only possible solution that can exist.
You are welcome to write to me, and ask more questions....
Barak
Co-Founder & CPO at Prevasio an AlgoSec Company
4 年Good post Barak. Interesting reading. Recently I read that TomTom has an HD-Map based tech, with sensors they call TomTom RoadDNA and they provide integration with AutoStream. https://www.tomtom.com/products/hd-map/