To be adopted in industry and on the roads, autonomous vehicles need to be perceptive and safe. What technologies will enable self-driving cars, forklifts, and drones to see and communicate accurately?
5D Robotics Inc. says it has the answer. Last month, it acquired Time Domain for its expertise in radio and radar sensors.
Carlsbad, Calif.-based 5D Robotics has developed guidance and positioning technologies used in 30 autonomous vehicles and robots. Its Behavior Engine is a reusable software suite, and its Adaptive Navigation Engine combines inputs from multiple sensor types.
“5D is at the unique intersection of accurate positioning and reliable behavior,” said David Bruemmer, CEO and founder of 5D Robotics. “We initially focused on behavior, but if you don’t have reliable positioning, you don’t have a functioning system.”
“It’s obvious even to people playing Pokemon Go,” he said. “But in robotics, it’s not the same problem as in the mobile/telecom world. We can’t deliver [a payload] if it’s off by a few centimeters or feet and get critical safety and efficiency benefits from autonomous forklifts or scissor lifts.”
Not only do automated guided vehicles in factories and warehouses use 5D’s systems, but so too can aerial drones and self-driving cars.
Time Domain sells the PulsOn modules, which are small, low-power, ultra-wideband (UWB) ranging radio and radar sensors. These modules are accurate to within 2cm and are useful for navigation indoors and as an alternative to GPS.
“Time Domain’s custom UWB chip is already embedded in our Universal Positioning Node and pairs with 5D’s Behavior Engine and Adaptive Positioning software,” Bruemmer told Robotics Business Review. “Together, we will significantly reduce the cost and size, enabling a new generation of assured mobility products.”
Tagging and mapping
“With Time Domain using UWB, ranging and radar technology is a universal module that we can insert into a wide range of robotics applications,” Bruemmer said. “We added our software into this module, and we can add it to everything, including drones, hard hats, and safety vests.”
“We’ve created an intelligent ecosystem, where we can track everything with tremendous precision,” he said. “We can guide motion, but with a human, we can do safety alerts and can perceive them where lasers could not.”
“5D and Time Domain provide mechanisms to rethink fundamental interactions between people and robots and between different types of robots,” Bruemmer said. “With only one solution, our tag works on a dog, a drone, or a tank.”
“We can get drones to follow people, drones to follow a truck,” he said. “A couple of big utilities have expressed interest for mapping and inspecting power lines — the concept of a virtual tether includes a UWB tag and a drone, without requiring GPS or a direct controller.”
Self-driving cars will need more than visual cameras, GPS maps, or lidar to be safe; they will need to be able to see objects in real time and communicate with one another and an enabled infrastructure, Bruemmer said.
“Can I drive a car on a premapped road in ideal conditions? Absolutely,” he said. “It has been a bit of a surprise to me to see how optimistic a lot of these folks have been,” he said.
“Let’s focus on building a P2P [peer-to-peer] positioning network,” he said. “Imagine a future where it’s not about autonomy — the car will figure it out — but where infrastructure and interactions are highly controlled in a framework. You can’t just slam a camera on a car and let it figure it all out.”
A path to safer autonomous vehicles
“The sensors in a cellphone aren’t enough to someday drive a car,” he said. “If a cellphone call drops, it’s not a huge problem, but for connected vehicles, you need good information for critical real-time decisions. The recent Tesla crash is a good example of that.”
“People don’t understand how vulnerable optics systems are right now. Both regular cameras and lidar can be stultified by bright sunlight,” explained Bruemmer. “The perceptual systems that robots have leaned on have problems with dynamic environments, and we’re trying to get reliability to 100 percent.”
“We’ve got to start doing [vehicle perception] differently,” Bruemmer said. “GPS isn’t going to get better or work in parking garages. Premapping is no good if people are moving around in a space.”
“Our implementation of UWB is unique,” Bruemmer said. “Our 87 patents cover a certain approach — pulsed-state UWB. It can reject multipath bounces at 350 meters and use the same ping to create a depth image.”
“You can have a panel or module, a phased array of antennas all sending out pings, like Knight Rider,” Bruemmer joked. “UWB sounds strange, but it works as a GPS enhancement.”
Safety and the ability to navigate in bad weather are important factors for outdoor autonomy.
“My work with the military for 10 years indicated that need,” Bruemmer said. “No one ever questions the need for precision when you’re dealing with land mines or power lines.”
“The automotive environment is just as complex as the military one — there are unexpected problems,” he noted. “We’re by no means the only ones to realize this. All of the major automakers promise full autonomy but realize that it’s not yet reliable. Some are coming to us and saying they need help with obscurants.”
5D Robotics and Time Domain’s technology allows self-driving cars to pick out the edges of curbs and be aware of obstacles beyond the line of sight.
“Unlike optics-based approaches, [5D’s systems] can see through a variety of surfaces, including car walls, drywall, vegetation, and some concrete,” Bruemmer said. “Not only can we range other active paths; we can also range to light posts, other cars, and bicycles.”
“By combining 5D with UWB, wideband with radar, we can provide a lidar replacement and P2P ranging,” he said. “This allows us to do powerful things.”
“It’s not as exciting as doing it [just] with a camera, but it’s actually way more valuable,” Bruemmer said. “A car could see the edge of a road in a blizzard, when optics don’t work.”
“We’re calling it ‘active positioning,'” he said. “UWB radar is different … range is more a means of anchoring down to situations where you have the luxury of [a 5D-tagged] piece of infrastructure. There are not many to start with, but you could build it into municipal lights. [Active positioning] can see at all times, even without tags.”
“If car companies want to increase the level of autonomy, they must think more at a systems level rather about individual vehicles,” Bruemmer said. “If we think about the problem right — can provide centimeter-level peer-to-peer positioning that can’t be hacked.”
IoT needs accurate positioning data
Bruemmer also asserted that his organization is ready for the Internet of Things (IoT), which promises to combine big data and artificial intelligence for business purposes.
“When we hear about IoT, it has science fiction-like goals, but basically, it all requires the same basic capability — centimeter-level positioning and a few degrees of accuracy in terms of heading,” he said. “There’s lots of crazy talk about IoT in industry, but you need accurate data and to get the benefits of identification.”
“To do anything regarding people’s locations in a car dealership, or to do more important things like keeping a forklift from hitting people around a blind corner, you need connected vehicles.”
“It’s also true for augmented reality — it all requires the same exact technology,” Bruemmer said. “When people talk about IoT — it requires what we’re doing.”
As robots become mobile sensors in factories, warehouses, hospitals, and more, they need both the ability to navigate and the ability to gather useful data for analysis and response.
“Many know where they want to get to, but they don’t know how to get there,” Bruemmer said. “They’ve created lots of fascinating apps in simulation, but they don’t know the real world. We’re moving from vagaries of GPS to precise, reliable, accurate positioning.”
More on Autonomous Vehicles:
- InVia Robotics Promises Easier Automated Materials Handling
- Report: Autonomous Vehicles Just Down the Road
- Automated Process Analytics Extend Productivity as IoT Approaches
- Machine Vision Attracts International Investment
- The Essential Interview: William ‘Red’ Whittaker, Field Robotics Pioneer and Entrepreneur
- Self-Driving Cars Get U.K. Government Support
- 5D Robotics and United Rentals Partner on Self-Driving Vehicles
Phased array could be a game changer
5D Robotics is ambitiously looking at applying its technology to heavy equipment, aerial drones, and car manufacturing.
“Our phased-array pulsed radar is a wholly new product and is going to be a game-changer for automotive, forklifts, etc.,” Bruemmer said. “This allows us to drastically reduce the costs of these modules.”
“We’re doing a lot of cross-pollination with existing customers,” Bruemmer said. “We’re also bringing more robotics capabilities to existing customers.”
“Every one of the automotive companies we’re talking with right now is interested,” he said. “Heavy equipment partners are extremely interested.”
“It’s also good for drones for deliveries from trucks,” he said. “Our systems can deal with clutter in the air, as well as people and vehicles, as long as UWB tags are up.”
In April, 5D Robotics acquired Aerial MOB LLC, which provides aerial cinematography for film and TV productions. Aerial MOB, which has offices in Los Angeles and San Diego, was the first company to get Federal Aviation Administration approval for filming with drones.
5D’s Aerial division will also provide mapping and multispectral imagery data for the utility and construction markets, among others.
5D Robotics has about 50 employees now, but it expects to grow, partly through partnerships.
“We’re not going to build drones or cars, but we’re proving out concepts, writing software, developing patents, and showing the world what works,” Bruemmer said. “Whether it’s for augmented reality, making sure that drones don’t land on the White House lawn, or smart transportation, they all require P2P positioning.”
“Our implementation of connected vehicle technology, which is fully functional right now, is the solution,” he said.