Toposens 3D Sonar Seeks Objects Overlooked by Lidar
A dream of robotic fish inspires inexpensive sensing technology for critical areas close to automated vehicles.
Alexander Rudoy was spending almost every hour of his free time developing a robotic toy fish that could sense its underwater surroundings. At the time he was studying electromechanical engineering at the Munich University of Applied Sciences, where he earned a master’s degree in 2015. Rudoy never finished his robo-guppy, but even before graduating he founded Toposens, a company that now utilizes the guppy’s underlying close-distance ultrasound technology for a wide range of vital automated-vehicle operations.
Rudoy and his partners at Toposens – managing director Tobias Bahnemann and head of development Rinaldo Perichini – have been promoting the need for a cost-effective near-distance sensor for vehicles and robots for about three years. The response from auto companies, which are primarily focusing on higher-speed, long-range automated functions, was muted.
Rudoy and Perichini previously worked at BMW on technologies ranging from automation to gesture controls. And the auto industry, now in the throes of solving for the full set of automated-driving operations, is finally listening. “Every autonomous car starts and stops, so it needs to perceive what is very close,” Bahnemann said. “You need centimeter-accurate, close-distance sensing or you could run over a dog or a child playing in the area. These are critical things that don’t happen 99.9 percent of the time, but if you’re not covered, it’s a big problem.”
Bahnemann admitted that a low-cost, low-power 3D ultrasonic sensor is “not as sexy” as lidar, camera-vision and radar sensors. But in many cases, those sensors leave a gap in coverage at the near-distance curb level. Automated vehicles need awareness of those areas to safely maneuver at low speeds while leaving or entering parking spots, for example.
Millions of vehicles already have ultrasonic parking sensors, which are combined with mandated backup cameras to warn drivers about obstacles and small objects directly around the car. But with that current one-dimensional sonar technology and no human driver behind the wheel, an autonomous vehicle – whether a sizeable people-moving shuttle or an icebox-sized delivery robot – lacks safety-critical levels of certainty.
Three microphones are listening
Perhaps not surprisingly, the key to Toposen’s technology (including the latest generation found in its TS3 3D sensor released in July 2019) is software. Rudoy and the company, now with 22 employees, continued to develop the algorithm since his student days. One of its U.S. patents was issued in March 2018 and another is pending. The Toposens sensor projects sound and then takes time-of-flight measurements from three microelectromechanical systems (MEMS) microphones. The hardware stack is comprised mostly of off-the-shelf components.
“It’s the software that does the magic and enables the 3D processing,” Bahnemann said. Those receivers are located close together in a 1-cm area. Based on its algorithms, a three-dimensional point-cloud of the near-distance surroundings including protruding and elevated objects is formed. The detection range is up to 5 meters (16.4 ft) at up to a 180-degree angle. The accuracy is approximately 1 cm up to 1 meter. The immediate surroundings are represented multiple times per second as coordinates in a Cartesian system.
Robotaxis from Waymo and others use lidar for this near-distance sensing, but the cost of multiple lidars can be prohibitive for lower levels of autonomy and mass-market advanced driver-assistance systems (ADAS) applications. Bahnemann said that Toposens 3D ultrasonic sensors, which are about the size of a matchbox, will likely sell for less than $20 when built at scale, allowing several of them to be used on a vehicle. He cited detection of curbs and walls, self-parking and precise vehicle positioning for inductive and robotic EV charging as promising automotive applications.
Toposens lists Daimler, Porsche, Audi and Continental as customers for its 3D sensor, which is not yet a mass-market product according to Bahnemann, and the company is currently seeking Series-A funding. Bahnemann said that competitors will eventually try to emulate the company’s 3D ultrasonic approach and he believes that Toposens has a head start of several years. This dates back to when co-founder Rudoy was dreaming about a robotic toy swimming around a small fish tank rather than millions of robotic vehicles driving around global cities.