From the Editor – Barry Manz
A long time ago, the 1950s to be exact, Greyhound got a lot of traction from its slogan “leave the driving to us”, so much so the company kept using it for 40 years. Today, the auto industry wants us to leave the driving to them, and to make that possible, vehicles will have to account for every conceivable eventuality and take action to keep us safe. That, needless to say, is a very tall order because there are an infinite number of these “eventualities”, and no single type of sensor can detect them all.
Consequently, the auto industry is using everything from cameras to radar, GPS, lidar, and maps, which makes for a complex, computationally intense network of sensors whose data must be fused and analyzed, after which a decision can be made and action taken. Unfortunately, even in reliable weather conditions, automotive researchers have found that the current state of ADAS systems will still experience some type of issue an average of every 8 mi. When this happens, they will (hopefully) turn back the driving to us, if we happen to be watching.
No Sensor Serves All
Cameras work by passively observing light reflected or emitted by the surface environment. While cameras enable a high-resolution view of the environment, changes that block light from reaching the camera will degrade or render them non-functional. These obscurations could be caused by sun glare, bright lights, a truck or object blocking the view, snow or ice on the lens, rain or snow in the air, drops of water on the windshield, or fog.
Lidar, which operates by sending out light and determining distance and intensity from the reflections, allows operation at night, detects objects, and can provide map-based positioning. However, in addition to being expensive, it is susceptible to many of the same issues as cameras.
Although automotive radar provides a lower resolution image of the environment than a camera, it has several advantages. Radar can transmit and receive through falling snow and rain with less loss, but current radar technologies are still limited in penetration depths and are sensitive to ice or snow build-up because they typically operate at millimeter-wave frequencies.
GPS can determine a vehicle’s position almost anywhere outdoors but is too inaccurate and unreliable for functions like lane keeping. It also fails when signals aren’t available, such as when a vehicle is next to tall buildings, in valleys, under overpasses, inside of parking garages, under tree cover, and in tunnels. It is also susceptible to reflections from nearby objects such as buildings, overhead road signs, and other vehicles. GPS signals are also very weak, which makes them susceptible to interference.
Overcoming Sensor Limitations
In the hope of remedying this situation, a company appropriately called GPR (formerly WaveSense) has developed a system that maps the substructure beneath a vehicle with ground-penetrating radar (Figure 1). The core GPR technology of the company’s product was developed by a team at MIT’s Lincoln Laboratory led by Byron Stanley, who left to co-found WaveSense with Tarik Bolat, formerly the executive vice president at Renewable Energy Trust, where he helped build one of the leading independent solar and wind power producers in North America. Building on the work at MIT, GPR has enhanced the technology’s capabilities, and in the latest version called Aegis, has reduced the size of the enclosure by 80%, making it small and thin enough to be mounted under most vehicles.
The system works by sending low-power VHF radar signals into the earth to generate a baseline map of a road’s subsurface that is stored in the system’s internal memory. The GPR’s main component is a waterproof closely-spaced 12-element antenna array. A single-board computer is used to correlate the GPR data into a three-dimensional GPS-tagged map and extract a corrected GPS position estimate. This map then becomes the reference for future travels over that stretch of road. As there is a reasonable possibility that something has changed since the last “visit,” the system can update its current data to reflect it. The relatively deep subsurface features are important because they are inherently stable and less susceptible to the dynamics of the world above.
As explained in an interview published in Wards Auto, Bolat says, “the first part is map-creating vehicles with sensors driving high-value routes such as major highways and high-traffic parking lots and garages. The second part is a crowd-sourcing element. Vehicles equipped with GPR will need more passes over an area to create a production-quality map, but as customer usage accelerates, vehicles will continuously expand the map. The more users we have, the better the product which attracts more users, and so on.” It will work with virtually any ground condition, from clean paved roads to those with snow, ice, water, mud, slush, sand, and salt (Figure 2), and can continuously enhance its performance and features through over-the-air upgrades.
Ground penetrating radar isn’t a new concept. The first patent, issued to Gotthelf Leimbach and Heinrich Löwy in 1910 (just 6 years after radar was first patented), explained the use CW radar to locate buried objects. A few years later, it was complemented by a patent exploring use of pulsed rather than CW signals, but it was decades before it began to take off, driven by military applications to detect mines and other munitions. It has since been used for studying bedrock, soil, groundwater, ice, the presence of diamonds, non-destructive testing, archeology, ground conditions on the Moon, and dozens more. But for the automotive industry, it’s brand new.
The system is not designed to become the only sensor required to satisfy all the capabilities required for ADAS or autonomous vehicles, but instead fuses its data with cameras, radar, and lidar. The fused and analyzed data is passed to vehicle control systems for use in automated vehicle control and reduces the need for continual modifications to high-resolution road maps. As the data from the other sensors is complementary it can be used to identify the surface conditions with more accuracy and robustness when the results are fused with those of the GPR system. The combination of these sensors, each with unique benefits, should be significantly enhanced by GPR, as it adds a new dimension that its counterparts cannot provide.
Controlling the vehicle can include controlling the velocity, acceleration, orientation, angular velocity and acceleration of the vehicle, which can be continuously controlled via one or more vehicle navigation commands to maintain the vehicle at the desired position along a trip path or to maintain the safety of the vehicle.
For example, the speed of the vehicle may be modified along segments of the path to maintain safe operation, to accommodate speed limits that can be affected by surface conditions, and to achieve a desired completion time for traversing the trip path. It may also include providing information to or assisting with traction control systems as well as wheel torque, with the driver warned to act as a result. In addition, when roads are repaved or resurfaced, the prior surface may be removed and new asphalt applied that will create differences in the radar signal reflections. GPR will detect these changes and add them to its data repository.
GPR says it has already mapped more than 25,000 miles of roads, its prototype systems are being tested by automakers, and the company is also working with suppliers and AV companies. The company has not limited its attention only to the auto industry as GPR may be useful by federal, state, and local governments for determining roads and bridges that require maintenance, as well as for indoor and underground navigation, such as in mining.