Accelerating Autonomy to Every Car

Accelerating Autonomy to Every Car

Lux Capital invests in Aeva, empowering machines to sense their surroundings as we do

Autonomous cars need finely-calibrated sensors to function safely in changing, unforgiving environments. In addition to being cheap and reliable, they must enable the vehicle to be deployable without being overseen by an army of PhD roboticists. As big sensor companies continue to make incrementally better sensors, the onus is upon a startup with the audacity to rethink how autonomous cars sense and perceive their surroundings.

As an engineer at General Motors, I witnessed the challenges suppliers face selling into the autonomous supply chain. Which is why, after a decade of having seen countless interesting ideas for radar, lidar, and and computer vision, we chose not to invest in any of them. Furthermore, the automotive supply chain has historically been a buzzsaw for startups: slow-moving, allergic to risk, and price-sensitive to the point where many new entrants wonder what possessed them to pursue it in the first place. Risk-aversion is to the extent where car companies prefer access to multiple vendors over whiz-bang technology. It’s little surprise that historically, manufacturers have relied on clever branding, marketing, incentives, and gimmicks to attract buyers into their showrooms.

However, times are changing for big car companies and their suppliers: Ride-sharing has gone from niche to mainstream, and automated cars have gone from possibility to inevitability. Automakers know that the status quo will move away from carmakers that are faster at slapping metal and plastic together, to those that offer quality, inexpensive transportation. The automotive business of the future is transitioning from moving metal to moving people, literally and metaphorically. In fear of being put in the same perilous position as their own suppliers by the likes of Uber and Zoox, conventional car companies are clamoring for the technology that would put them in a position of strength. Meanwhile, big players in other industries, such as Google, Baidu, and Tencent, are also exploring getting into the transportation game. They are accelerating design cycles, taking risks on new technology, and putting their balance sheets to work.

When we were introduced to Aeva cofounder Soroush Salehian by our friend Tommy Leep, we took the meeting to help an aspiring young entrepreneur. Expecting to give advice, and perhaps even talk him out of it, we found our skepticism being challenged by the vision (no pun intended) for Aeva; to bring perception to robots as WiFi chips brought connectivity to everyday products. Soroush and his cofounder Mina Rezk proposed a new type of sensor that immerses robots in a way that is not possible with conventional sensing technology.

Aeva has delivered a single sensor that empowers a machine to perceive its surroundings, like humans do, but with far greater fidelity. Aeva is simplifying autonomy by putting more perception into the sensor, with the goal of making “autonomy” as simple an add-on as Broadcom and Qualcomm have done with “internet connectivity” for portable devices.

Live screen shots of an Aeva sensor mounted to a vehicle in Mountain View, CA. The top-left box displays the conventional range information beyond 200m which is greater than conventional lidar. The bottom-left box displays reflectance which is orders of magnitude higher-fidelity than the best pulse lidar. The top-right window highlights the magic Aeva brings to bear: static objects (i.e., trees, the road, parked cars, and standing pedestrians) are shown in white, separated from moving objects, whose velocity is indicated by color, e.g., towards the red end of the color spectrum for objects moving away and blue for oncoming objects. Note that the system automatically distinguishes between cars waiting to turn left and oncoming traffic from behind.

Aeva does more than just collapse a few sensors into one: it relaxes the requirements for all sensors and simplifies the autonomy stack as a whole. Furthermore, Aeva eliminates the multi-sensor calibration challenge that remains unsolved today. Aeva enables an unprecedented level of perception by not only natively measuring how surrounding objects are moving, but potentially predicting their trajectories.

Venture capitalists, seizing the opportunity, have forked over hundreds of millions of dollars to fund companies that build better lidar, radar, computer vision, maps, tele-operations, and simulators; all of these offer point solutions to the very complex problem. Unfortunately the boundaries of each solution are not well-understood. Startups are still figuring out the nature of the system they will be integrated into, as Waymo, Cruise, Zoox, Aurora, Argo, and other leaders may be forging their own paths with their own proprietary architectures. It could turn out that many of these new gadget and software licensing companies are end up no different from their predecessors selling brake lining and infotainment systems. Perhaps others, like Aeva, by offering a modular solution that simplify big problems, can emerge as dominant companies in the space.

Over the past 18 months, Soroush and Mina have assembled a cognitively diverse, world-class team of all ages, technical backgrounds, and life experiences. Aeva is partnered with key automotive and ride-sharing companies that represent many millions of units per year. We are proud to be partnered with Soroush and Mina on their journey as they grow a magical organization making impossible robots, possible.

We’re delighted to be partnered with Soroush (right) and Mina (left) on their journey from the founding of Aeva, to building a fantastic team, to demonstrating their first working sensor, to shipping them to major carmakers and ride-sharing companies to accelerate their autonomous ambitions.
written by
Shahin Farshchi, PhD
General Partner

Inspired by Knight Rider and Star Trek, Shahin grew up with a passion to endow superpowers to humanity through feats of engineering. He learned BASIC on an IBM PC XT clone he built at his aunt’s computer store and used to dial into Bulletin Board Systems when he was in the 4th grade. He learned about engines by taking apart an Alfa Romeo at his Uncle’s repair shop a few years later. He aspired to design microchips, software, and systems that could someday amount to the fictional K.I.T.T., and build warp drives that could propel humanity to far corners of the galaxy.

Shahin has had the privilege of partnering with amazing founding teams for over two decades. After working for several software startups, he built his first startup in 2004 building upon his PhD research designing chips, systems, and software to capture and interpret brain signals. He later had the pleasure of meeting Lux’s founding team, who invited him to partner with them in 2006, where he has cofounded and led Lux’s investments into companies that have gone on to become publicly traded (e.g., NYSE:AEVA and NYSE:PL), and acquired by the likes of Intel (Nervana), Amazon (Zoox), Silicon Laboratories (Silicon Clocks), and Lattice Semiconductor (SiBeam).

written by

Accelerating Autonomy to Every Car

visit source
Text Link