Ouster to Challenge Waymo with Digital LiDAR

Article By : Junko Yoshida

While the automotive lidar market remains in flux, Ouster, a San Francisco-based lidar startup founded in 2016, is confident that the company can beat Waymo at its own game in broad-based, non-automotive markets...

With the automotive lidar market in flux, lidar vendors are hunting for ancillary markets. Waymo is taking its home-grown lidar to cast a wider net beyond automotive, where it’s strong, into lidar-hungry markets that range from robotics to surveillance and agriculture. Ouster, a San Francisco-based lidar startup founded in 2016, is confident it can beat Waymo at its own game in non-automotive markets.

How Ouster’s digital lidar works.

The upstart Ouster has a variety of lidar models designed to meet the needs of multiple markets. In contrast, Waymo is plunging into the vast non-automotive market with just one model, Honeycomb.

In an interview with EE Times last week, Ouster’s founder and CEO Angus Pacala boasted that his company has already picked up 700 design wins over 15 different industries in 50 countries.

Impressive, but where’s Outster’s advantage?

Pacala said, “We chose technology designed to work in many markets.” Ouster has developed a lidar platform built on “all-CMOS semiconductors.” That makes Ouster’s products “digital lidars,” according to Pacala.

Ouster’s competitors, including Velodyne and Waymo, deploy hundreds of off-the-shelf discrete components to make their spinning lidars work. In contrast, Ouster has developed tightly integrated custom vertical cavity surface emitting lasers (VCSELs) and another ASIC that incorporates single photon avalanche diodes (SPADs) arrays.

Ouster’s platform also includes Xilinx’s FPGA, responsible for processing massive amount of data.

Ouster (Source: Ouster)

This is essentially the same digital lidar architecture — VCSEL and SPAD arrays — Apple used for its iPad Pro 11, explained Pacala.


Recommended
Look Inside iPad Pro 11’s LiDAR Scanner

SJ Consulting Report: Transformation in Retail, Logistics and Transportation


Alexis Debray, technology and market analyst at Yole Développement, told EE Times that while Waymo’s real-world testing experience in autonomous vehicles would likely advantage its lidar in the automotive market, Waymo lacks experience in serving hundreds of different clients in the industrial market. In Debray’s view, the key to winning multiple markets is to offer “many models” and “customization.”  “For example, Ouster offers around 50 different models, all based on the same technology but with various ranges, fields of view, and resolutions,” Debray observed.

A variety of lidar technologies
Today’s lidars have a plethora of options for directing laser beams. These include spinning lidar, mechanical scanning lidar, optical phased array lidar and flash lidar.

There is no agreement on a single method within in the industry about which technology is better. As Pacala wrote in his own blog:

… there are multiple spinning lidar companies using side emitter diode lasers and APDs, multiple FMCW/doppler lidar companies using telecom style fiber optics and frequency tunable lasers, multiple 1550 nm galvo scanning lidar companies using fiber lasers, and multiple MEMS scanning lidar companies…

Aurora, just last week, announced the debut, on the Medium, of its first light lidar as a sensor on Aurora’s next-generation test vehicles. The so-called light lidar is based on the FMCW lidar technology bequeathed to Aurora through its acquisition of Blackmore last year.

This sounds ground-breaking. But Pacala, asked about Aurora’s new FMCW lidar, told EE Times, “That lidar is designed for super-specific applications – like directly looking down the road.” It’s most likely that Aurora’s autonomous vehicles will also use Ouster’s lidars, he added.

Among these competitors, Pacala stressed, “Ouster is the first company to commercialize a high performance SPAD and VCSEL approach.”

Next up: Solid-state multi-beam flash lidars
To be clear, the lidars Ouster has been selling thus far are all spinning units. The startup, however, revealed plans to roll out solid-state lidars in “the next couple of years,” according to Pacala.

Ouster does not regard as a huge hurdle the transition from today’s spinning lidars to flash lidars. This is because Outster plans to use “the exact same technology [VCSEL and SPAD] for solid-state flash lidars.

Ouster isn’t revealing details about its solid-state flash lidars. But compared to currently available flash lidars, which typically illuminate the entire field with a single flash, Ouster is designing multi-laser flash systems with an array of thousands — or millions — of lasers.

Pacala is promising it to be “the first performant, low-cost, true solid-state lidar” on the market.

Customers and partners
Asked to name its customers and partners, Pacala cited Nvidia. Ouster is working with Nvidia to provide lidar sensors for use in Level 3 to Level 5 autonomous driving systems being developed by major global OEMs targeting production in 2022. Pacala added that Postmates has selected the Ouster OS1 lidar sensor for use in its autonomous delivery rover deploying in Los Angeles. Ike Robotics is another partner. Ike chose Ouster’s OS1 lidar sensor for its commercial trucking platform. Separately, Uber recently announced the acquisition of Postmates.

The price range of Ouster’s lidars spans from $1,000 to $20,000, according to Pacala.

In contrast, Waymo, which announced 16 months ago its plan to pursue non-automotive segments by licensing its internally developed lidar, recently confirmed multiple design wins. However, it offered no specifics on its technology, pricing and partners.

Lidar that can function much like an image sensor
During the interview, Ouster’s CEO repeatedly mentioned his quest for a sensor design that unites in one device the best aspects of lidars and cameras.

Pacala’s motivation is the realization that deep-learning research for cameras is far more advanced than that of lidars. The deep learning algorithms developed for camera images don’t translate to the 3D point clouds generated by lidars.

With that in mind, Ouster has developed firmware updates to its lidar, so that its OS1, for example, now “outputs fixed resolution depth images, signal images, and ambient images in real time, all without a camera,” according to the startup. “The data layers are perfectly spatially correlated, with zero temporal mismatch or shutter effects, and have 16 bits per pixel and linear photo response.”

The video clip above shows both point clouds and the camera like output of the sensor. Ouster’s OS2-128, mounted on top of the vehicle just in front of the sunroof, drives on 101 south through San Francisco. The OS2’s 200 meter range and 128 channels of resolution enables long-distance object detection

In short, Ouster is already building two-dimensional arrays of lasers and detectors similar to the image sensor in a camera. “We’ve already proven [that] out with our spinning lidars,” Pacala confirmed. Ousters spinning lidars are “already densely packed 2D arrays, though we still scan them to achieve 360. We are now working to expand their size considerably to reach wide field of view in an all solid-state device,” he explained. The goal for Ouster is to create a new flash lidar class in which each “pixel” is served by its own dedicated laser-detector pair.

Advantages of digital lidar
Problems with analog lidars are many. For example, adding more channels requires additional physical components, which increases potential failures. An increase in resolution also directly leads to a hike in cost.

In contrast, with digital lidars, Ouster expects a clear path to exponential growth in resolution and pixel counts in lidar, similar to the impact silicon has had on CMOS image sensors. For 256-channel digital lidar, for example,” no change is necessary in architecture, size, cost, complexity, reliability,” according to Ouster. The company just needs to “upgrade to  swap in newer ASIC/laser chip.”

As lidar resolution improves, the processor in a lidar system must be able to keep up with massive amounts of data.

Willard Tu, senior director of automotive at Xilinx, told EE Times, “That’s where FPGA comes in.” Fixed-function silicon cannot possibly meet the flexibility and adaptability required by Ouster’s evolving lidars and its further improvements.  The combination of the Xilinx programmable logic, advanced DSP, and connectivity options makes the Xilinx platform apart from competition,” Tu explained. This frees customers like Ouster from the chore of redesigning an entire lidar architecture.

Another edge is that Xilinx has a long history in functional safety. The company can assure its customers that Xilinx automotive-qualified products offer them the reliability they need, Tu explained.

Leave a comment