Apple was serious about partnering with BMW for Project Titan in 2017- maybe because Steve Jobs, its iconic founder, loved his BMW. Which LiDAR company will invoke similar passions for the I-Car as it starts gathering steam once again? Given Apple’s pioneering efforts in integrating LiDAR into personal computing and smartphones, it is interesting to explore their potential choices for the I-Car.
The requirements for an I-Car LiDAR (I-LiDAR) are likely to be centered around performance, reliability, data quality and monetization, styling and vehicle integration. Cost is not likely to be a major driver.
The LiDAR currently deployed in Apple consumer devices use 9XX nm VCSELs (Vertical Cavity Surface Emitting Lasers) and SPADs (Single Photon Avalanche Photodiodes). It is a flash LiDAR device with limited resolution, Field of View (FoV) and range – great for indoor mapping and photography, but not really deployable for a safety application in an autonomous vehicle. The architecture is compelling from a styling and size perspective – the absence of messy mechanical scanners enables sleek integration, without the LiDAR appearing prominent or visible. Extending the performance and reliability of this architecture into automotive applications like L3 (conditional automation) and L4 (full automation in a prescribed operational domain) was discussed in a recent article. Similar to Apple, Sense Photonics, Opsys, ZF-Ibeo and Ouster use VCSELs and SPADs in the 8XX-9XX nm wavelengths, but extend range, FoV and PPS (points/second, similar to resolution in a camera) performance. Sense’s LiDAR uses no scanning (flash LiDAR), whereas Ouster, ZF and Opsys deploy electronic scanning with no mechanically moving parts. The absence of mechanical scanners enhances reliability and enables graceful vehicle integration – although the range and PPS performance of 9XX nm systems is likely to be more limited than 14XX-15XX nm systems because of eye-safety constraints.
It is entirely possible, of course, that Apple decides to develop its own I-LiDAR. If it does, the odds are that it will leverage the experience from the I-Phone and I-Pad LiDAR industrialization, and pursue the VCSEL-SPAD design. Partnerships or other relationships with the companies listed in the previous paragraph is a possibility. Ideally, a flash LiDAR is desirable, although electronic scanning is a reasonable alternate choice. The reasons for wanting to develop the LiDAR in-house are similar to other autonomous vehicle pioneers like Waymo, Aurora, Argo, Yandex and Cruise. These companies decided that deep integration of the LiDAR was critical for mapping and proper functioning of the perception and machine learning stacks, and invested heavily in acquisitions or internal developments to master this critical sensor technology. It is no coincidence that they are the leaders in L4 vehicle autonomy. Having the right LiDAR is certainly a critical enabler for this new mode of transportation.
The BMW angle would suggest a LiDAR that uses MEMs scanning of an edge emitting laser at 9XX nm and a classical linear mode avalanche photodiode based receiver. Many companies are developing LiDARs based on this architecture, but Israel based Innoviz would seem to be a favorite in this category given that they are designed into the BMW platform already through Magna, a Tier 1 supplier. However, the primary factor guiding this LiDAR choice is the automotive partner itself – it still remains an open question as to whether Apple will team up with an automotive OEM or Tier 1, and if so, what the specific partner choice will be.
Frequency Modulated Continuous Wave (FMCW) LiDAR at the 15XX nm wavelengths could be another option for the iCar. It offers an attractive roadmap for eventually realizing a chip scale LiDAR in a silicon photonics platform – however, the challenges are significant and it will take a fair amount of time, luck and money to get there. Aeva’s LiDAR could potentially be a choice – both co-founders are ex-Apple employees, and the company recently announced two high profile ex-employees to their board – Steve Zadesky and Alex Hitzinger (who currently heads Artemis, Volkswagen’s Autonomous Vehicle effort). It is likely that in the near term, these LiDARs will still require opto-mechanical scanning and an optical fiber feed to transmit and receive photons from the FoV to the data processing engine. The immaturity of the chip scale realization may not be an immediate issue for Apple’s I-Car efforts since it is not as schedule constrained as other companies relying on external funding with the accompanying urgency to prove their business models quickly.
Another potential play could be Argo’s recently announced LiDAR that provides a rotating surround view with 360 degree HFOV coverage and uses 14XX-15XX nm SPADs. It appears to be a reasonable choice from a performance perspective (range of 400 m, good resolution, etc.). However, the visual effect of a large roof mounted rotating device look is probably too jarring from a style perspective for a company like Apple to deploy. Other potential 15XX nm solutions include Luminar, Aeye and Innovusion – all use lower sensitivity photodetectors than SPADs but compensate by using more powerful fiber lasers (which unfortunately are more expensive, bulky and less amenable to automation and high volume scaling that semiconductor based solutions can attain).
The decision eventually will come down to the applications the I-Car is focused on. If operation in urban areas is important, the 9XX nm VCSEL-SPAD solution is the likely choice. Highway applications and the eventual possibility of a chip scale LiDAR would push the needle towards 15XX nm FMCW LiDAR. High resolution mapping and collection of monetizable data would probably drive Apple to develop the LiDAR in-house using the VCSEL-SPAD architecture it has already mastered for consumer devices. Given the varied applications and needs, it is likely that multiple LiDAR platforms will be needed.
One thing is certain. Similar to the iPhone, we probably will not know where the I-LiDAR is headed – until it is happens!