Talk To Us
Radically accelerate your roadmap with Tangram Vision's perception tools and infrastructure.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Radically accelerate your roadmap with Tangram Vision's perception tools and infrastructure.
Table of Contents
Spring is here, which means it’s time for our updated Perception Sensor Industry Map — now in its fourth year (wow, really?!). This year, we've dropped the word "sensor" from the title, and going forward, it will be known as the Perception Industry Map. This is to more accurately reflect the fact that the pace of new hardware/sensor entrants has slowed, while software has accelerated and is now the fastest growing source of new perception map companies. That’s not to say that there haven’t been major shifts in other sectors of the map, however. There’s been quite a lot of flux between 2023 and 2024, which we’ll explain in detail below.
While some companies have exited sectors (for instance Waymo ceasing to sell its Laserbear Honeycomb LiDAR to third parties, and Huawei generally ceding Chinese sensor markets to Hesai and RoboSense), very few companies have gone under. This is particularly surprising given the general decrease in capital availability in a post-zero-interest-rate (aka “ZIRP”) world. Perhaps many of these companies raised war chests before markets tightened, or perhaps the uplift in deployments for robots and AVs generally has buoyed revenues. Because of the private nature of most of these companies, it’s hard to know where they stand in terms of cash position and revenue. All we can say is that it’s 2024, and they’re still here.
For the 2024 edition of the market map, we’ve decided to make new distinctions in sectors that have proliferated the most over the past few years. Thus, 3D sensing is now broken up into 3D sensing and industrial 3D sensing. Cameras have been split into embeddable and industrial cameras. And what was once strictly event cameras is now “emerging technology”. The latter reflects quite a bit of hardware and software innovation representing technologies moving from the lab and entering real world use. Will all of these companies be here next year? Subscribe to our newsletter and be notified when the 2025 Perception Industry Map arrives and see!
Download the map as a PDF here.
So, with that out of the way, let’s track the changes!
In our biased opinion, the biggest news in 3D sensing is the arrival of HiFi, Tangram Vision’s own depth sensor. In all seriousness, however, the 3D sensing market has seen low levels of truly new product introduction over the past couple of years, with a stable set of options from market participants like Intel RealSense, Luxonis, Stereolabs, and Orbbec. That is changing in 2024. Not only will Tangram Vision HiFi be shipping, but Stereolabs has launched their new Zed X, Orbbec has launched a slew of new Gemini models, Luxonis will be shipping their RVC4-equipped depth sensors, and Intel RealSense has doubled down on biometric use cases for their 2024 sensor updates. Thankfully, the market for depth sensors is also growing by leaps and bounds, meaning that all of these new entrants should benefit from scaled deployments and new users in 2024 and beyond.
While these sensors were grouped with other 3D sensors last year, we felt it appropriate to segment them into their own category this year. This is because these sensors compete primarily for a different set of customers and markets. You’d be hard pressed to find a Photoneo or Ensenso depth sensor on a mobile robot, but they are pervasive in industrial use cases like traditional industrial automation. Because of this, they are typically orders of magnitude more expensive than the 3D sensors in our other 3D sensing category. This category has remained relatively stable, with few new market entrants or product introductions since it emerged in the late 2010s.
Given their ubiquity in automotive applications, the ultrasonic sector is dominated by large OEM suppliers like Continental, Texas Instruments, and Bosch. However, there is still innovation at the early-stage level, and this year’s map sees the addition of Sonair. Sonair’s upcoming sensor uses a MEMS transducer that transmits a high-frequency ultrasonic signal which is then decoded on reflection by a set of proprietary algorithms. According to their website, they’ll be shipping their first prototype units this year. We’re excited to see (or is it hear?) them in action!
No other perception sensor category has seen more boom and bust than LiDAR. In the Scanning category, we see that Huawei has exited the commercial market (along with their solid state lines), and most robotics companies are now standardizing on Hesai or Ouster for 3D LiDAR, and SICK or Hokuyo for 2D LiDAR. This year’s map adds two lower cost dev-oriented options - SlamTec (a long time participant) and Unitree (new, and no doubt a result of their expanding success in deploying their humanoid and animal form-factor robots).
Additions to this year’s map include Seyond (formerly Innovusion), and PreAct. Both have been around for quite some time, and both are now starting to commercialize their offerings. In particular, PreAct is notable for delivering a flash LiDAR sensor for under $400. For those companies that once relied on the Intel RealSense L515, the PreAct Mojave represents an interesting replacement with a similar performance profile.
What we don’t see in LiDAR markets yet is any sort of massive, scaled adoption. Market leaders like Ouster/Veloydne and Luminar still have yet to find high-growth customers with deployments in the hundreds of thousands and beyond. Even passenger vehicle deployments such as Valeo’s Scala LiDAR in Mercedes Benz’ DRIVE PILOT system, the Hesai AT128 in the new Lotus Eletre, and Luminar’s delayed, yet upcoming debut in the Volvo EX90 aren’t expected to drive massive scale
While the participants in thermal sensing have remained stable since last year’s market map, we can confidently say from our own interactions with robot and autonomous vehicle companies that thermal sensing is one of the most intriguing modalities for the next generation of autonomy. In particular, the ability to merge thermal sensing with standard RGB or mono cameras adds a perception backstop in low- to no-light conditions, and occluded conditions like dense fog. We anticipate that this trend will continue, giving the favorable price per unit for these sensors versus competing modalities like LiDAR and radar.
Four or so years ago, mmWave and 4D radar were all the rage. Now, it seems that investors have cooled on funding new radar startups, as scaled demand has yet to materialize. In the meantime, large OEMs like Continental and Bosch have been building increasing market share with automotive companies to power L1/L2/L3 ADAS systems. Of note: after having sworn off radar to focus strictly on camera-based ADAS, Tesla has chosen to reintroduce radar into the Model X and Model S. This is a high-resolution radar of their own design, but it has not yet been activated. If so, it seems likely that it will be for discrete capabilities such as forward collision avoidance.
As we’ve noted over the past few years, quite a few robotics companies are choosing to “roll their own” depth systems. When they do, they usually select an option from one of these manufacturers. For our own HiFi sensor, we’ve chosen a set of optics from ams OSRAM. For those robotics and AV companies that operate primarily in outdoor spaces, another rising trend is the deployment of high dynamic range (or “HDR”) cameras that can operate more effectively on varied lighting conditions, from very dark to very bright.
Just like with industrial 3D sensors, the industrial camera (aka “machine vision”) sector has remained highly stable over the past decade. Due to their high cost and integration features focused more on industrial automation, these industrial cameras are rarely found on mobile robots or autonomous vehicles, but can be found as accessories to pick-and-place robots and other fixed position factory robots.
We’re going to need to do something about this category. If there’s one takeaway from this year’s map, it’s that this sector is exploding with activity, both from new entrants, as well as participants that have reached levels of maturity that merit inclusion. In next year’s market map, we’ll likely split these out into subcategories, such as: sensor fusion, SLAM, AI/ML training, and general purpose perception platforms. In this year’s map, we see expansion in the sensor fusion and AI/ML categories.
We believe that as robotics and autonomy companies mature and move towards deploying at scale, much of the perception tooling they implemented during development will become a limiting factor — and this includes homegrown tools. As a result, perception software companies (ourselves included) are now seeing a decay of “not invented here” approaches at robotics and AV companies, and a much greater willingness to use third party software to solve complex perception tasks.
In the world of autonomy, IMUs can be coarsely segmented into expensive, higher accuracy all-in-one designs, and lower-cost, lower-accuracy chipset IMUs. Of note in the former category is the continued evolution of Movella, the growing IMU conglomerate that was once mCube, Xsens and Kinduct. As a merged entity, they now possess the scale and product scope to compete against heavyweight incumbents such as Honeywell and Lord Microstrain.
From the market map’s inception until last year, this category was dedicated to neuromorphic and event cameras. Unfortunately, this excluded a host of fascinating technologies that have been gaining interest from robotics and autonomy engineers. Hence, the new moniker of “Emerging Technology”, and the inclusion of interesting new players like Vayu Robotics (polarized long-range depth sensing), Opteran (insect-inspired high-speed depth imaging), and DreamVu (mirror-based 360° robot vision).
As we’ve seen with neuromorphic cameras, it can take quite some time for these emerging technologies to gain acceptance for deployment in production. There is often a catch 22 where these new technologies arrive with high prices due to a lack of a scaled supply chain. Yet to achieve the adoption for a scaled supply chain, prices must come down.
And a quick note on neuromorphic cameras: one of the market leaders, Inivation, has been acquired by neuromorpic software provider SynSense to create a joint software and hardware company driving further market adoption for event cameras.
And there we have it: our 2024 Perception Market Map. From an initial 75 companies in 2021, it has now grown to include 105 companies, an increase of 40%. Many sectors (like LiDAR and thermal cameras) have stabilized, but some sectors (like software and emerging technologies) have seen quite a bit of new company creation and growth. Overall, however, the picture is one of continued expansion as the key markets served by perception (robotics, autonomous vehicles, and industrial automation) themselves continue to expand.
Did we miss your company? Let us know, and we’ll be certain to include it on next year’s map. Download the map as a PDF here.