Crystal Group helping pioneer vehicles of tomorrow today

When the COVID-19 pandemic shut down conferences and industry events, VSI Labs, a leading researcher of active safety and autonomous vehicle (AV) technologies, refused to hide away the latest industry advances in a garage. Instead, the Minnesota-based testing organization took to the nation’s highways, launching a series of cross-country drives to showcase its state-of-the-art autonomous test vehicles, highlight various vendor technologies and collect data for those same partners on performance under different conditions and driving environments. Last week, VSI’s latest 2,500-plus mile road show – the Automated Trip South from Minneapolis to Austin, Texas and back – brought two specially outfitted Ford Fusion Titanium hybrids to the Corridor, offering a peek at the future of driverless automation. The “beating hearts” of the vehicles? The AVC0161 autonomous vehicle computer manufactured at Crystal Group’s Hiawatha headquarters, which is being deployed in defense, industrial and commercial settings as part of a futuristic – but coming – driverless future. Founded in 1987, Crystal Group is a maker of rugged high-performing computer hardware that delivers real-time, seamless, secure data processing, storage and transmission in what the company calls “unpredictable, unmanned and unforgiving environments.” According to Brian Hamed, Crystal Group’s Industrial Program manager, the high-end computing device found a following in the defense, industrial and commercial space thanks to its liquid-cooled ability to support the rapid processing units needed to capture and prioritize massive amounts of real-time data recorded by onboard sensors, radar, LiDAR (Light Detection and Ranging) and cameras. The computer allows for 360-degree situational awareness and split-second decision making in automated vehicles. “One thing we’re trying to do is grow our autonomous business. It’s pretty exciting to see it in a vehicle, actually,” he said of the test drive, one of several VSI Labs has taken throughout the pandemic. “To actually see the drive, to see some of the different databases and systems that hold information you can see on their website… it’s really interesting to see and exciting for us, as a company and as employees who helped build it and support that equipment.” Several dozen Crystal Group employees came out last week to see the result of their efforts in a test car full of autonomous vehicle innovations. In addition to Crystal Group’s ruggedized computer, VSI Lab’s current demonstration vehicle features RTK Navigation from OXTS, LiDAR from Ouster, thermal cameras from FLIR, signal phase & timing from Bolton & Menk, simulation by Siemens, tire grip indicators from Nira Dynamics and HD maps from HERE HD Live Map, to name just a few. The vehicle – not quite capable of driverless action yet – offers adaptive cruise control that adjusts speed to maintain a safe distance between cars; path following, which synchronizes localization data with a target path or HD map to calculate trajectory, acceleration and deceleration; and automatic emergency braking that engages immediately when it detects a pedestrian. It also features map-based lane changing, INS localization, point cloud localization, vision- and map-based lane keeping, highway interchanging via high-definition maps, and object detection, tracking and avoidance to avoid collisions. What does it mean? “What they’re trying to do is replace the human driver,” said Micah Snodgrass, technical director at Crystal Group, adding vehicles can’t do that – at least, not yet. “Nobody’s 100% of the way there yet, right?  So it’s taking over most of the normal boring driving function. Think if you were a driver’s ed instructor. You still have the main driver there with their foot ready to hit the brake pedal. But you have a teenager sitting there controlling the vehicle under normal, non-life threatening, non-weird, non-bad-weather situations. So, it’s watching the road, noticing lane markers. The computer might not be able to handle some construction zones. But for the most part, it’ll take a normal trip from point A to point B as long as there’s nothing unusual.” John Cooper, vice president of business development for VSI Labs, said his company’s cross-country test drives aren’t conducted in service of building the perfect automated vehicle. It’s all about testing various systems on VSI’s open architecture. “It allows us to plug and unplug LiDAR X for LIDAR Y, camera A for camera B, radar X  for radar Y,” he explained. “We have a real open architecture, and the core of that architecture right now is, we’re happy to say, is the Crystal rugged machine … We just know when we plug it in. It works.” Phil Magney, founder and president of VSI Labs, said using Crystal Group’s computer as a base to test driverless vehicle functions for its coast-to-coast awareness campaigns was a natural. “We need stuff that is really rugged and can handle the harsh conditions,” he said. “When it’s 95 degrees, and we’re at the test track, and we’re doing emergency braking all day long, and just slamming on the brakes … time after time, after time, any other computer, any typical computer would fail under those conditions. This computer is very, very ruggedized in its ability to handle shock, vibration and massive temperature fluctuations. It’s a water-cooled system, so it’s probably the best computer platform you could have for doing R&D of these technologies. And it’s got a ton of computing power in it, it’s got dozens of processors and tons of memory and storage and that sort of stuff. It’s really core to the whole system.” Crystal’s potential Crystal Group’s AVC0161 was designed to combine high-powered computing, data handling and storage with rugged, compact design that can withstand harsh conditions and environments, including potholes, collisions and temperature extremes that can cause traditional systems to fail. The high-performance onboard computers feature Intel Xeon processors, up to 2 TB of DDR4 memory and sophisticated power and thermal management stabilized in a size, weight and power (SWaP)-optimized aluminum chassis. The unit is in use in several military and industry contexts for defense, industrial, critical infrastructure and commercial applications – including trucking companies and robotics developers, which have partnered with the company to accelerate the development and performance of autonomous trucks and heavy construction machinery. In March, Crystal Group announced a collaboration with California-based TuSimple, which has built the world’s first Autonomous Freight Network (AFN), an integrated software and hardware solution that allows for Level 4 autonomous driving for semis. The Society of Automotive Engineers defines 6 levels of driving automation, ranging from 0 (fully manual) to 5 (fully autonomous). That means TuSimple’s fleet has full self-driving capabilities with some limitations on driving locations or conditions. The custom autonomous vehicle computer Crystal developed for TuSimple features two GPUs to process the vast amount of real-time data captured by an autonomous truck’s network of sensors, radar and cameras, and provides a full view of the operating environment for safe, successful driverless navigation. It can manage significant heat generated by extreme processing power, while withstanding the thermal, shock and fluctuations conditions common in long-haul trucking. “Ongoing, candid collaboration with TuSimple resulted in a custom solution with enhanced processing power, thermal management and shock absorption that not only met their functional requirements, but also raised industry benchmarks for safety, efficiency and reliability,” Mr. Snodgrass said. Crystal and VSI officials said that widespread adoption of AI-enabled autonomous capabilities is expected to explode over the next five years, particularly in the trucking and construction industries, bringing better operating efficiencies, reduced operating costs and safer roadways and work sites that are less prone to human error. The self-driving truck market is projected to reach $1.67 billion by 2025, according to Crystal Group estimates, while autonomous construction is expected to reach $14 billion by 2023. When do we get to 5? So how far away are we from a future where we can theoretically sip coffee and read the newspaper while our cars escort us effortlessly to work? Not as far as you might think. If today’s average driver-operated car is a level 0 or 1, Mr. Snodgrass said, readily available technologies like lane departure systems, which warn the driver when the vehicle begins to move out of its lane, bring vehicles up to level 2 or 3. “Between three and four, the driver can stop touching the steering wheel,” he explained. “At five, they can just take the steering wheel out and throw it away.” “We’re really close now,” Mr. Magney said. “The technologies are already in the market right now, with all these safety systems that are coming in on the late model cars – lane keep assist and automated parking or emergency braking – those are really kind of the building blocks of automated driving. So, it’s not something that’s setting itself up for, you know, 10 or 20 years from now. It’s here and now.” Mr. Magney said the most advanced commercially available cars today are at level 2, allowing drivers to, for example, press a button ordering the vehicle to automatically drive within the context of the lane. Cars from Tesla, Cadillac, BMW, Subaru and others already offer that feature. “Now, the high-end stuff, the level 4 stuff – the robo-taxis, the driverless trucks – those are being deployed now on a test basis in Arizona and some parts of Texas. So they’re coming, and they’re going to come to major metro areas [first],” he continued, adding it would probably be five years plus before that technology made its way north to the Midwest. “But over the next two, three years, you’re going to start seeing more and more of that emerge – your Ubers and your Lyfts and stuff like that.” Mr. Snodgrass said true level 5 automation has numerous kinks to iron out in the meantime, with even vaunted Tesla models struggling on the true open road. “For short, controlled areas like a parking lot, or Disneyland, or a convention center, college campus, those sort of controlled environments … that whole robo-taxi idea works very well,” he said. “And over the road trucking is going in the same direction … But when you get into decision making areas, should I stop for gas here? That’s where that’s where the computers aren’t quite there yet. There’s the trucking people, there’s the delivery people, there’s the grocery people, there’s the vehicles that take little kids back and forth to school, so many things like that. And you don’t have an all-in-one solution yet.” In addition, Mr. Hamed said the regulatory environment for driverless vehicles is still unclear and industry standards remain somewhat murky. “What [the environment] should look like is something a lot of regulatory agencies are going through and trying to determine,” he said, noting that rules of the road in more autonomous vehicle-friendly states like Iowa, Kansas, Texas and California differ from one another and differ again from the other 46 states. “And you have not only the rules over here, but in Europe, China and places like that. All these autonomous vehicles are starting to pop up as well as all these different software companies that don’t all so much understand the regulations or standards for automotive, but they have automotive customers who understand the technical field, and they’re trying to wrap their head around computer science. So it’s an interesting, dynamic … a lot of experts are focusing on their area of expertise and not everybody’s talking.”