English
Connected autonomous vehicles must always be able to calculate their own location. If they cannot, they will be unable to maintain the right position on the road, avoid obstacles or navigate safely to a destination.
Positioning is therefore a critical capability, and in an autonomous vehicle it’s generated by combining data from a variety of sensors – from satellite receivers (GNSS) to lidar, radar, cameras, inertial measurement units (IMUs) and cellular technologies like WiFi, 4G/5G and LTE.
GNSS is key to autonomous vehicle positioning
Of these sensors, only one – GNSS – can generate an absolute position. Global navigation satellite systems like GPS, GLONASS, Galileo and BeiDou enable the vehicle to understand where on the Earth’s surface it is; its precise latitude and longitude. All others can only generate the position of the vehicle relative to previous known positions, or to other objects in the environment, like roadside infrastructure, cell towers and other vehicles.
That makes GNSS an essential sensor to guide navigation, as well as a source of positioning truth that data from other sensors can be compared with.
GNSS also plays another vital – and lesser known – role. Atomic clocks on global navigation satellites continuously communicate a time signal accurate to within 100 nanoseconds. Those precise timestamps can be used to determine the exact time something happened in or to the vehicle, which will be critical for vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications.
Developers must understand the impact of GNSS signal degradation
The criticality of GNSS as a sensor within the connected autonomous vehicle means developers need to have a solid understanding of the many environmental factors that can disrupt, distort or block satellite signals. Two essential factors to understand are obscuration and multipath.
Obscuration refers to satellite signals, which work on a line of sight basis, being blocked by objects such as buildings, hillsides and dense foliage. A GNSS receiver needs to be able to receive signals from four satellites in order to generate a position, but in urban canyons surrounded by tall buildings, fewer than four may be in view at any one time. And even when four are visible, the dilution of precision (DoP) resulting from their proximity to each other may cause an inaccurate position to be calculated.
Multipath refers to a satellite signal bouncing off the surface of something in the vehicle’s environment – like a tall building, high-sided vehicle, hillside or even the ground. This reflected signal has slightly further to travel, and so arrives at the receiver slightly later than line-of-sight signals. Without proper mitigation, this can cause the receiver to output an inaccurate position. (However, multipath signals can also be useful as they can reach receivers in areas where line of sight signals are obscured.)
For these reasons, CAV developers and researchers have been looking for ways to incorporate realistic multipath and obscuration simulation into the test lab.
Historically, this has been done using statistical models of the likely arrival point of GNSS signals in different environments. That’s been acceptable while humans have been in control of navigation, but as vehicles become increasingly autonomous, it’s necessary to get a detailed understanding of the effects of obscuration and multipath on the vehicle’s ability to generate an accurate position.
3D simulation offers a good solution
Spirent has been working closely with the industry to help developers of connected and autonomous vehicles simulate the effect of obscuration and multipath in a realistic way.
The solution we’ve developed in partnership with Oktal Synthetic Environment allows researchers to load realistic 3D models of different driving environments, including elements like vehicles and pedestrians that may further contribute to RF signal degradation. It also allows these models to be shared with a driving simulator.
The effects of obscuration and multipath on the GNSS signal are accurately modelled in real time – even taking into account the properties of different physical materials – so the impact on the GNSS receiver and the vehicle’s sensor fusion algorithms can be realistically evaluated.
Researchers can choose from a range of geo-typical environments, such as urban canyons or tree-lined highways, or commission Spirent to build a geo-specific model of a real place. Geo-specific models developed to date include areas of San Francisco and San Jose, as well as the campus of Warwick University, which houses one of the UK’s leading R&D centres for intelligent vehicles.
Realistic simulation will drive critical design decisions
The ability to realistically simulate multipath and obscuration brings a host of benefits to researchers and developers of autonomous vehicle control systems. It can guide critical design decisions like where to place the GNSS antenna on the vehicle; which type of GNSS receiver to use; and when to hand off to other position sensors if the GNSS signal becomes too degraded.
To learn more about how Spirent can help, contact us to discuss your simulation requirements.
