Researchers at the University of California have developed a positioning system which promises greater accuracy for drones and self-driving vehicles, by utilising near signals such as cellular and Wi-Fi data instead of GPS.
The team, from the ASPIN Laboratory at UC Riverside, presented its paper at the 2016 Institute of Navigation Global Navigation Satellite System Conference in Portland, Oregon, in September. The work centres around the possibilities for Signals of Opportunity (SOPs) to augment Inertial Navigation Systems (INSs) when GPS data drops out or loses resolution.
Precision GPS navigation is a symbiosis between GPS and INS, the latter of which relies on dead reckoning and just-in-time advance appraisal, when unaided. When GPS signal does degrade, the INS can diverge, potentially dangerously, from the true mathematical GPS course described for a journey. Using ambient data as local references creates a useful navigational resource that sits between the dumb dead reckoning of INS and a more accurate GPS signal which is, however, too distant, and may be subject to casual interference.
The team used the concept to map the accuracy of a drone flight over lower Los Angeles. Here the red line represents the ideal route line, the fragmented yellow indicates the actual GPS data received, and the blue line indicates the more accurate routing made possible when SOP data is added to the GPS signal:
Common vehicle navigation systems make use of the orbital Global Navigation Satellite System (GNSS), which is comprised of international systems including the Russian GLONASS, China’s Beidou, Europe’s Galileo system and the United States’ GPS, but the resolution available at general commercial and consumer levels from this approach are, the researchers contend, inadequate for the more critical emerging drone and self-driving technologies.
For one thing, line-of-sight is an issue, with abutments and other occlusions degrading guidance accuracy; on a more sinister note, GPS signals are unencrypted and susceptible to hacking, for which reason Lidar has proved of interest to the manufacturers of drones and autonomous vehicles.
Zak Kassas, the assistant professor of electrical and computer engineering at the University’s Bourns College of Engineering, believes this massively redundant approach could be improved:
“By adding more and more sensors, researchers are throwing ‘everything but the kitchen sink’ to prepare autonomous vehicle navigation systems for the inevitable scenario that GPS signals become unavailable. We took a different approach, which is to exploit signals that are already out there in the environment.”
In one sense the team seems to be abandoning one set of ‘kitchen sink’ approaches (Lidar et al) to increasing accuracy for another, insofar as a multiplicity of varied sources are likely to provide the best reference information; this is because the clock states and position of the SOPs are not necessarily known in advance of the journey, and may vary in quality or become unavailable depending on the journey’s length. The team notes that this facet of the scheme presents a similar challenge to the SLAM problem in robotics.