Autonomous Vehicles Driven by Photons


Journal of Selected Topics in Quantum Electronics doi: 10.1109/JSTQE.2022.3162577 

Since the invention of the automobile back in the late 19th century, advances in what is incorporated into a car today have been tremendous. Broadly speaking, in the last twenty years the proliferation of several types of motors has captured our attention, departing from the original steam powered to electric, internal combustion and more recently hydrogen-based engines. In the last decade we’ve been also been flashed by the possibility of cars driving alone, in a manner such as the 1980’s iconic TV show Knight Rider, having had its 35-year anniversary in 2022.

 While modern vehicles improve since more than two decades ago, technologies such as radars that help us gather the surroundings are only suited for specific situations like parking. However, driving alone in the most demanding conditions, that is at road/highway speed in whatever the weather and traffic conditions are, requires the combination of several high-performance technologies.

Among those technologies, the “eyes” of the car are key and how the scene ahead is evaluated in terms of objects present, capturing their relative distances and speeds. Photons, the fundamental particle that hits our retina and allows us to visually interact with the world, are the response. Light detection and ranging (Lidar) are the analog to radar but using photons. A Lidar system will emit photons, such as the sun does, and will collect those scattered back, similarly to what human eyes do.

One could think primarily on two approaches for Lidar. In the first, the system flashes the scene ahead with photons, and then has an array of detectors, such as insect ommatidia based eyes. In a second, the Lidar emits narrow beams of photons, and steers them to scan the scene, like how the old cathode-ray tube TV systems worked. Then, a single detector is employed to collect back-scattered photons. For any of those two Lidar vision techniques, several approaches for signal generation, detection and processing are possible too, mostly borrowed from radar. In one, pulses can be emitted at a given time, and distance to objects is then determined by their relative time of return. Others can use a more sophisticated approach in which frequency modulated signals allow for determining not only the relative position, but the relative velocity between the Lidar equipped vehicle and any other in the scene. The consideration of all the above, in a scenario where other Lidars and interfering signals (such as sunlight, or even objects very close shadowing other further away below the detection limit) entails engineering the system departing from the specific application; the requirements are different for robots driving at modest speeds in a storage, where surroundings do not change suddenly, than for the dreamt planted Knight Rider scenes.

Such Lidar systems can be built using off-the-shelf photonic devices, but broadly speaking with luck it would use space like a shoe box. Lidars are to be boarded in vehicles subject to movement and vibrations. Furthermore, vehicles will certainly equip more than one Lidar for several reasons. But it is reasonable to consider at least four of those systems in a car. And cars are manufactured in humongous quantities. Thus, for Lidars to be in car they need to be based on chips, in particular photonic chips, also known as photonic integrated circuits (PICs).

 In a recent published paper in IEEE Journal of Selected Topics in Quantum Electronics, the authors proposed a step towards a Lidar chip on a car, with architecture that can be scaled up to meet actual driving conditions. The photonic chip presented (see figure) can produce beams of photons in the near-infrared and steer them along the scene by means of a tunable laser diode. The chip architecture incorporates switches that allows for addressing a selected part of the scene (random access). The photonic chip can also adopt different resolutions to image different parts with coarser or finer detail (adjustable resolution). Finally, the architecture can be configured to outline large objects (object framing).

The photonic chip was manufactured in state-of-the-art CMOS-compatible foundry, using silicon-based technology. Hence, the device reported is amenable for mass manufacturing, a very relevant aspect to meet automotive market expected demand for Lidars. However, the new work highlighting silicon needs to be complemented with other materials to be able to meet all application requirements. In particular, the reported Lidar chip has a limited reach estimated of a few meters, while most demanding scenarios such as highway driving require reach of a few hundred meters. Silicon needs then to be complemented with III-V materials for such purpose (producing more photonics, thus extending reach in Lidar applications). PIC technologies, based on silicon, III-V semiconductors and other materials, and their combination (so-called hybrid integration) are maturing at a pace where Lidars in cars could be a reality by 2030. This work outlined and detailed the proposed hybrid integration to incorporate such optical amplifiers to the Lidar chip, and how the reach could be then extended to meet the demanding scenarios.

P. Muñoz et al., “Scalable Switched Slab Coupler Based Optical Phased Array on Silicon Nitride,” in IEEE Journal of Selected Topics in Quantum Electronics, vol. 28, no. 5: Lidars and Photonic Radars, pp. 1-16, Sept.-Oct. 2022, Art no. 8300416, doi: 10.1109/JSTQE.2022.3162577.