Cheap version of LiDAR is available, costing only 1% of Google's driverless radar

    Although LiDAR has been used in many smart cars, its bulky size and high cost are affordable for Google's recently released "Little Bean" driverless car. It seems that no "stupid" car company is willing to invest in this technology in its mass-produced cars. Therefore, if the size can be smaller and the price can be lower, such equipment may also enter the homes of ordinary people in the future.

    Behnam Behroozpour, a graduate student at the University of California, Berkeley, is currently working on a project to package a full 3D imaging system, commonly known as LIDAR, into a solid-state package that can be easily used on a smartphone or game controller.

    Behnam Behroozpour explains in detail for me:

    According to current ideas, such a lidar device consists of three parts. We first prepare to make a 3mm×3mm optoelectronic chip, then a CMOS sensor chip, and finally a chip-sized VCSEL (vertical cavity surface laser emitter) device. However, we have only tested the performance of individual chips separately before, and now we are starting to assemble them. The final experimental product is about the size of Microsoft's Kinect box.

Multi-pixel integrated 3-D camera using FMCW laser source

    Although it is not clear how accurate this integrated lidar system can achieve, according to current experimental data, the detectable range of the laser beam is about 10 meters. However, this is obviously far from enough for use in the automotive field, and this geek team from Berkeley has begun to study how to make it reach a range of 30m-100m.

    LIDAR measures the distance of a target object by emitting a beam and calculating the time required to receive the reflection. To prevent measurement errors caused by receiving surrounding reflected light, you can systematically adjust the energy value of the laser beam; you can even change its frequency, which makes it easier to obtain accurate data. Frequency adjustment like this is achieved through a very small, vibrating MEMS mirror placed on the right side of the silicon chip.

    The experimenters, led by Behnam Behroozpour, used electronic frequency modulation to make these mirror devices vibrate naturally. According to him, although this approach is more complicated and time-consuming, it can reduce the loss of light energy on the one hand and obtain an ideal signal-to-noise ratio on the other. The noise in this experiment mainly comes from diffuse light, which may be mistaken for the reflection of the laser beam to a large extent, thus causing interference in the measurement of experimental data.

Schematic diagram of hybrid 3D circuit of MEMS-electronic-optical chip

    The self-driving cars that are now available on the market rely on a large number of sensor technologies, including laser, radar, ultrasound, standard cameras, stereo cameras, GPS, inertial guidance and even Wi-Fi, and radar equipment is the most important part. On the one hand, radar can be used normally in any lighting and weather conditions; on the other hand, thanks to the high development of the aviation industry, its size and cost have been greatly reduced. Even if some cheap LIDAR devices are placed in inconspicuous places around the car body, they can still achieve high-resolution and accurate information scanning.

    But to achieve this, we have to reduce the cost first. The LIDAR device installed on the roof of Google's driverless car is said to cost $70,000. Behroozpour said he hopes that the first integrated LIDAR device he and his team designed can keep the cost to a few hundred dollars. Although this may still be expensive for smartphone and Xbox game enthusiasts, it is obviously very cost-effective to be able to achieve this level of a solid-state integrated system.

summary:

    Undoubtedly, if Behnam Behroozpour's wish can be realized in the future, then this small and low-cost lidar device can not only be used in driverless cars, but also hidden in your smartphone and your Kinect game controller. But there is a problem that cannot be ignored: such a laser box composed of three 3mm×3mm chips is fine for small devices, but if such a small gadget is used in a car carrying passengers, whether safety can be effectively guaranteed is still unknown.

    Take the comparison between SLR and compact cameras, compact cameras are thin and light, easy to carry, but they still cannot reach the image quality and lens range of SLR. However, the essence of innovation lies in breaking and establishing. The LIDAR used by Google's driverless cars is too expensive, which is indeed a big obstacle to its commercialization. Assuming that Behnam's research can be better in scanning object accuracy, range and physical images, then the spring of Google's driverless cars will come!