Can the GNSS Connection Solve Fully Autonomous Driving Challenges?

The jury is still out on whether GNSS can be an effective tool in driverless cars due to issues regarding multiple communications protocols, longer-range accuracy, and cost.


Roger Allan, Contributing Editor

For a fully autonomous vehicle to become a marketable reality, it must solve very difficult RF communications problems associated with the vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2X and V2I) global navigation satellite system (GNSS). Sensor-to-cloud-based AI wireless communications issues also must be addressed. These are mighty challenges that impact a range of hardware and software issues, and not everyone in the automotive, AI, sensor, and related sectors believes are easily achievable.

GNSS plays a critical role in next-generation positioning systems as the only source of absolute position, velocity, and time. However, it can only be accurate to within two to five meters for the position of the road. Next-generation fully autonomous vehicles will require a GNSS with centimeter-level (about 10 cm) accuracy, and must be able to operate with vehicles in all driving conditions.

Fully autonomous vehicles will be using dedicated short range communications (DSRC) for communicating V2V and V2X entities. This presents tough design challenges due to environmental variations affecting roadway vehicles that may be in rural, urban, and highway areas.

A contentious issue that has yet to be resolved is the communications protocols for the licensed intelligent transportation system ITS-G5, which is based on the IEEE 802.11P Wi-Fi standard, and the CV2X that’s based on the 3GPP standard. These two protocols provide two different communication access layers. The former is a designed with vehicle safety in mind, while the latter is smartphone cellular-based for direct connections between vehicles. The IEEE 8902.11P is an improved amendment to the IEEE80211 standard that adds wireless access in a vehicular environment between vehicles and roadside entities.

Many automakers favor the ITS-G5 approach, as does Autotalks, a pioneer in developing V2X technology. Others like BMW (one of the inventors of V2X technology) as well as Qualcomm favor the CV2X method. Arguments center on bandwidth use efficiency, interference, performance, and cost.

Line-of-sight and non-line-of-sight situations in which vehicles are communicating with one another via RF may be encountered. The distances between cars, their relative speeds, and the angles between approaching vehicles are directly responsible for signal loss, signal delay, Doppler radar shift, weather and lighting conditions, and other vehicle communications. Heavy rains, snow, and freezing environments can make sensors unable to detect fading roadway driving lane lines.


Cost-Effectiveness is Key

Issues regarding cost are major factors in this arena. Designers must be able to simultaneously integrate safety, performance, and complexity without any tradeoffs in each, which affects hardware and software designs. This is a big challenge for IC chipmakers, where software updates, changes, and maintenance require more computational power per chip to lower implementation costs. Most argue that cloud-based will be very helpful.

Donald Walker, CEO of Magna, cautions that “AI needs more refinement and is not there yet.” He questions whether or not AI can distinguish between a ball and a child chasing it on the road. He feels that autonomous features will depend on consumer demand, legislation, and lawsuits, all of which may make fully autonomous driving a long way off.

Attaining reliable conclusions from cloud-based data may not be practical for autonomous vehicles that will wirelessly communicate with one another. That’s what Matt Grob, Executive Vice President of technology at Qualcomm, believes: “In many cases, inference running entirely in the cloud will have issues for real-time applications that are latency-time-sensitive and mission-critical.”

Chris Osterwood, CTO of Carnegie Robotics, and Fergus Noble, CTO of Swift Navigation, have shown that a GNSS coupled with an inertial measurement unit (IMU) provides the foundation for an accurate centimeter-level GNSS-IMU platform for precise global navigation of driverless, fully autonomous cars. The satellite(s) provide the global-positioning information, while the IMU provides heading, pitch, and roll information.

Swift offers a MEMS-based Piksi Multi GNSS receiver that can be embedded in autonomous cars that can receive data from multiple satellites. A ruggedized version, the Duro, can be housed outdoors and attached to a structure or tower (Fig. 1).

Figure 1: A ruggedized version of Swift’s MEMS-based Piksi Multi GNSS receiver, called the Duro, can be housed outdoors and attached to a structure or tower.

The Autonomous Push Forward

Nevertheless, automakers and their software, AI, and sensor partners are racing to achieve a driverless fully autonomous vehicle. One of the biggest carmakers, GM, will conduct fully autonomous vehicle tests on an SAE Level 4 high-automation Chevy Bolt next April in a designated zone in the southern part of New York City’s Manhattan borough. This level calls for a driverless vehicle that performs all aspects of the dynamic driving task, even if a human driver doesn’t respond appropriately to a request to intervene.

Last year, GM acquired Cruise Automation, a startup working to transform traditional cars into self-driving vehicles using aftermarket hardware. GM has indicated that additional partnerships and/or acquisitions may follow to achieve the fully autonomous vehicle goal. And French automaker Groupe PSA and Hungary’s Almotive are jointly launching the second phase of an SAE Level 4 autonomous vehicle test at cruising speeds of 81 mph along 186 miles of French highways. Testing will be performed over a variety of weather, traffic, and lighting conditions.

Toyota said it will begin testing self-driving cars around 2020. And The BMW Group, Intel, and Mobileye have been joined by Canada’s Magna International, to develop a self-driving platform for driverless vehicles.

Mercedes-Benz International believes it may have an answer to a fully autonomous car. It laid out a scenario involving a sophisticated V2V and V2X approach that allows a car to see hazards before they’re perceived by the driver, and warns the driver and other road users in sufficient time. Such a platform will reportedly be unveiled very soon (Fig. 2).


Figure 2: Mercedes-Benz International plans to unveil a platform that enables a car to perceive hazards before they’re seen by the driver, and in turn warn the driver and other road users.

Start typing and press Enter to search