Page 12 - EE Times Europe Magazine – June 2024
P. 12

12 EE|Times EUROPE

        The Autonomous Car Industry in 2024: Sensors, Software and Safety


        high-reflectivity objects and low-reflectivity   radar could determine whether something    cars and the emergence of software-defined
        objects simultaneously. Consider, for example,   is metallic (high reflectivity) or organic (low   vehicles, a modern high-end electric car will
        a small child standing near a car. The reflection   reflectivity). With this distinction, radar   have more in common with a computer than
        from the car is orders of magnitude brighter   would become an even more complete sensor,   with an average car from 20 years ago.
        than that from the child. For the radar, detect-  offering low cost, operation under all weather
        ing the child is like trying photograph a grain   conditions, accurate ranging and velocity   SOFTWARE
        of rice in front of the sun. It’s very easy for the   and, finally, accurate object classification.   Horsepower is becoming democratized. Scal-
        child’s reflection to be lost to noise.   Researchers are working on intensity detec-  ing an electric motor to double or triple the
          Advances in semiconductor technology and   tion capabilities now, and the technology   horsepower is much cheaper than doubling
        radar design are improving radar’s dynamic   could soon make its way onto radar or, alter-  the output of an internal combustion engine.
        range. For example, at CES 2024, Mobileye   natively, onto the car’s central autonomous   Cars that would put the supercars of yester-
        showed an example of a radar that could   driving and sensor fusion computer.    year to shame are now entering the market
        detect a wooden pallet next to a metal railing   Previously, radar data was processed on   at less than US$50,000. In combination with
        at nearly 240 meters.               the radar itself. Local processing makes sense   this, electric cars do not possess the differ-
                                            for a task like automatic emergency braking,   entiating characteristics of engines. They do
        It is an exciting time as           which requires only a flag that something is   not have turbo lag or differing exhaust notes;
        the automotive industry             stopped in front of the vehicle. But radar data   most do not even have more than one gear. In
                                                                                short, future cars will be differentiable not by
                                            is so rich that startups like Zendar, Zadar Labs
        goes through the most               and others want to handle the processing on a   their performance but by their software.
                                                                                  Software-defined vehicles, connectivity
                                            much more powerful, central computer. Such
        transformative journey in its       machines not only would handle radar data   and autonomy also provide a route to perfect
        history.                            but would accept high-resolution video feeds   safety. The problem with sensors, even the
                                            from multiple cameras, LiDAR data and more.
                                                                                best and most advanced sensors, is that
                                              With so much data to process, the industry   they cannot see around corners or through
          Processing is another area in which radar   is moving to more advanced semiconductors   buildings. No matter how many radar, LiDAR
        utility can be improved. The industry has   and more powerful systems-on-chip. Nvidia’s   or camera units are fitted to a car, they cannot
        started using the nomenclature “4D imag-  Thor, expected in 2025, is the prime example.   see that a speeding car on an adjacent city
        ing radar,” referring to azimuth (horizontal   With 2,000 TOPS of processing power, and   street is about to jump a red light. But con-
        angle), elevation (vertical angle), distance   utilizing TSMC’s 4-nm node technology, this   nected vehicles could.
        and relative velocity. But radar has another   SoC has more in common with a high-end   Connectivity and software offer a pathway
        dimension that could further improve its   gaming PC than with the chips running sys-  to collective perception, in which cars and
        performance: intensity, or reflection strength.   tems in most consumer cars today. With the   infrastructure would share their environmen-
        By considering the strength of a reflection,   advent of Chat GPT, the demand for AI within   tal knowledge with each other. Armed with









































        Radar being used for adaptive cruise control (Source: IDTechEx)

        JUNE 2024 | www.eetimes.eu
   7   8   9   10   11   12   13   14   15   16   17