There Must Be 860 Ways To Build A LiDAR Sensor For Autonomous Vehicles — Part 1
DARPA started it all. With the Grand Challenge a decade ago — with researchers competing to drive robotic cars through unpredictable terrain, with planned obstacles, and as fast as possible — without a human driver.
The Grand Challenge catalyzed commercial activity to develop automated driving for regular roads. A decade later, there is significant progress. The initial euphoria of the research phase is transitioning into the sober reality of managing investor expectations, cash flow, operations, product development and solid customer acquisitions.
A clear winner in all of this is the intense focus on sensing and perception, and the role of LiDAR as one of the key enablers of the autonomy utopia. LiDAR enables 3D sensing with high resolution — something other sensors like 2D cameras and radar cannot do. Most researchers in the field believe that LiDAR is critical in order to ensure obstacle avoidance, localization and navigation. Velodyne, a maker of acoustic speakers developed the first LiDARs for the DARPA Grand Challenge — this experience positioned them as one of the leading LiDAR companies today.
Optics emerged as a critical technology two decades ago with the adoption of optical networks for communications, data, video and high-speed internet. Ancillary markets in medicine, industrial sensing and materials processing benefited from investments in telecommunications. Similarly, the prospect of “opticalizing” future cars is providing a second birth and excitement in the optics industry. LiDAR, which originally was focused on defense and space applications promised the potential of becoming a mainstream fixture, like cameras are today in cell phones and cars. The enabling potential of LiDAR for autonomous vehicles has attracted upwards of $1.5B from venture investments. Similar levels of investments have occurred within corporations like OEMs, Tier 1 suppliers and technology companies (General Motors, Toyota, Bosch, Continental, Waymo, Argo and Aurora).
Over 80 venture funded LiDAR companies exist today, initially focused on automotive LIDAR. The slowdown in the AV revolution [discussed by Richard Bishop in a recent Forbes.com article, and recent announcements from Waymo, Daimler, etc.] and the limited customer traction that many of these companies have in the automotive space, has forced them to pivot to other applications like security, aerial mapping, drones and industrial automation. It is encouraging that other applications are being spawned, similar to the optics-telecom boom two decades ago. But it is clear that all these companies will not survive on a stand-alone basis– the eventual pie is just not big enough, and there is not enough “intelligent” money from venture funds and corporations to fund all these companies through the next stage of industrialization. My estimate is that no more than 10 companies will survive to service the ADAS and AV markets.
Apart from the euphoria, there are reasons why so many companies could even convince investors that they should exist. Considering the various design choices available on wavelength (4), LiDAR type (3), laser types (3), detector types (4) and scanning approaches (6) yields 864 combinations. With other technological options like 3D point cloud processing approaches and capabilities like event-based LiDAR, results in even more combinatorial options. Not all are physically practical or make sense, but certainly at least ½ are, which explains why VCs chose to invest in so many companies (for example, Velodyne, Quanergy, Innoviz, Luminar, Aeye, Aeva, Ouster, etc.). They backed one or many of these horses in the initial phases and now need to figure out how to protect or leverage their investments gracefully.
ohn Dexheimer is one of the few investment bankers in the optical space who has experienced the boom and bust cycles for the optical industry during the telecom and AV periods. I asked him to compare these two industries: “I saw analogies for automotive LiDAR to the optical telecom ramp. Both have seen the entry of large and mid-sized firms, and floods of venture and hedge fund money for start-ups. A big difference is there were clear, immediate economic drivers for the telecom boom, while in AV and ADAS the economics are less clear near term, with design-experiment choices being made based on perceived end-user needs and desires”.
Telecom had very rigid and structured standards for products entering the pipeline, whereas LiDAR in AVs is not yet there. John Dexheimer has been focusing on standards for AV LIDAR and raises the following question: “What are the effective standards bodies for auto doing to help narrow the acceptable testing metrics so that volumes can be achieved beyond trials and limited geo-fenced applications? “
When and how will the AV LIDAR space consolidate? Strong players with significant funding, execution teams, and best-in-class customer and supplier relationships will survive. But the question is which weak players will get bought out by large Tier 1 suppliers or tech companies, or merge with the stronger, independent LiDAR companies. Without such moves, companies face bankruptcy or IP sale scenarios. Part 2 of this article will explore this — primarily along the idea that the post processing software stacks for perception and sensor fusion, as well as the supply chain aspects across these 864 design combinations will be significant drivers. But this is not the whole story — there is great promise in some of the 864 combinations which have not yet been funded !
Stay tuned for Part 2.
Follow me on LinkedIn.
Sabbir Rangwala, is a Board of Advisor for AutoM8 ( A Fountech Ventures portfolio company). In the past he has led the automotive LIDAR business at Princeton Lightwave until 2017. Currently the Founder at Patience Consulting, the company provides expertise on AVs, perception and LiDAR. Patiently!
Fountech Ventures - Venture Builders of Deep Tech AI Startups | LinkedIn
Fountech Ventures - Venture Builders of Deep Tech AI Startups | 312 followers on LinkedIn. We lovingly house, rapidly…