CRII: SaTC: Cyber Resilient Localization and Navigation for Autonomous Vehicles

Autonomous vehicles (AVs) in all modes of transportation (be it surface, aviation, or maritime) require accurate and reliable localization and navigation services in order to perform their autonomous functions. In many cases, a Global Positioning System (GPS) or Global Navigation Satellite System (GNSS) provides the required localization and navigation capabilities for surface transportation.

A GPS (or GNSS) mostly depends on satellites and radio communication, which are subject to various obstructions, such as high-rise buildings, walls and ceilings in garages and tunnels, and even thick cloud cover. Besides these factors, GPS devices are also subject to intentional threats, such as radio interference, spoofing on communications, data manipulation on transmitted messages, the jamming of GPS receiver channels, and disruptions to the GPS infrastructures.

Existing cyber-resilient solutions against GPS interference, such as high-definition maps and Wi-Fi or cellular-based technologies, are either data-intensive and costly or susceptible to signal interference and limited by coverage area. Alternatively, low-cost in-vehicle sensors (including gyroscopes, accelerometers, steering angle sensors, inertial measurement units, odometers, and cameras), which are not susceptible to signal interference, can provide effective strategies for detecting threats as well as locating and navigating AVs.

To establish cyber-resilient localization and navigation for autonomous driving in a roadway environment, the goal of this research is to develop sensor fusion approaches that combine outputs from low-cost in-vehicle sensors with those of onboard geographical information systems.

This proposed research addresses intentional and unintentional interference issues of GPS services in order to improve the operational safety of AVs. The project will first investigate and develop an approach to detect GPS interference (such as jamming, spoofing, and natural interference) by predicting and identifying vehicle states (including distance traveled, turning, and lane-change maneuvers) based on data from low-cost in-vehicle sensors and the onboard graphical information system.

This initial research thrust adopts a three-step approach: (i) create different types of jamming and spoofing attacks using interference devices in both a laboratory and controlled real-world environment; (ii) generate attack and attack-free datasets in real-world scenarios for predicting and detecting vehicle state information; and (iii) develop a deep sensor fusion model and dynamic time warping algorithm to detect GPS interference using the generated attack and attack-free datasets. The second stage of the project will develop an integrated cyber-resilient navigation system using data from in-vehicle sensors and the onboard graphical information system to guide an AV towards its intended destination in a GPS-denied or GPS-compromised environment.

This will involve a three-step process: (i) generate route creation and location information using the graphical information system; (ii) fuse steering angle and gyroscope data to identify lane changes and turning movements using deep fusion and dynamic time warping techniques developed in the first research thrust; and (iii) integrate graphical information and sensor-fusion data for localization and navigation when the GPS signal is compromised. The final stage of research will develop a proof-of-concept of the cyber-resilient localization and navigation system using a conventional vehicle with low-cost AV sensors.

Broader impacts entail a transformation of autonomous vehicle operations under intentional and unintentional interference.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.