Pervasive, Accurate and Reliable Location Based Services for Emergency Responders
Location based services (LBS) for GPS-denied environments have gained significant maturity in recent years. Although they have started being used in commercial environments, there are still several challenges that prevent their immediate adoption in emergency scenarios. For example, emergency scenarios preclude the use of LBS that require intensive survey or infrastructure deployment. Yet, the requirements for coverage and availability are paramount; it is not acceptable to have blind areas where people cannot be positioned, or their positions cannot be communicated to the incident commander. The requirements for reliability and accuracy are also particularly strict - sub 3-meter accuracy (95% of the time). Floorplan information may not always be available, and even if it is, it may have changed as a result of the incident. Finally, adverse conditions like smoke may hinder the use of some sensors (e.g. cameras).
The high level aim of the proposed project is to develop fit-for-purpose location based services for emergency responders, addressing the unique challenges discussed above. Our specific objectives are to develop: 1) pervasive LBS systems that work everywhere - from high rise buildings to deep basements, and from steel/concrete frame structures to smaller wood framed or load bearing masonry buildings; 2) accurate LBS that can locate responders in 3D with meter level accuracy; and 3) reliable LBS that can ensure such accuracy can be achieved even in adverse conditions (smoke, fire, wall collapse) and is reliably communicated to the incident commander with low latency.
We plan to adopt three distinct approaches to addressing these objectives. For pervasiveness, we advocate using inherently portable technologies, such as inertial tracking and vision-based egocentric localisation, which require no prior survey or deployment. For accuracy, we propose novel fusion methods that play on the strengths of each technology, and can mitigate the impact of noisy sensors and challenging building layouts. For reliability, we propose the use of a novel robust communication and positioning modality - low frequency magnetic fields - that can reach responders even if they are trapped under thick layers of rubble. We further boost reliability by designing novel adaptive techniques that can dynamically change sensor sampling to respond to changing conditions and dynamically evolving user requirements.
Selected Publications
-
Snoopy: Sniffing Your Smartwatches Passwords via Deep Sequence Learning
Chris Xiaoxuan Lu‚ Bowen Du‚ Hongkai Wen‚ Sen Wang‚ Andrew Markham‚ Ivan Martinovic‚ Yiran Shen and Niki Trigoni
In ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp). 2018.
Details about Snoopy: Sniffing Your Smartwatches Passwords via Deep Sequence Learning | BibTeX data for Snoopy: Sniffing Your Smartwatches Passwords via Deep Sequence Learning | Download snoopy_sg.key of Snoopy: Sniffing Your Smartwatches Passwords via Deep Sequence Learning | Download snoopy.pdf of Snoopy: Sniffing Your Smartwatches Passwords via Deep Sequence Learning
-
IONet: Learning to Cure the Curse of Drift in Inertial Odometry
Changhao Chen‚ Chris Xiaoxuan Lu‚ Andrew Markham and Niki Trigoni
In The Thirty−Second AAAI Conference on Artificial Intelligence (AAAI−18). 2018.
Details about IONet: Learning to Cure the Curse of Drift in Inertial Odometry | BibTeX data for IONet: Learning to Cure the Curse of Drift in Inertial Odometry | Download (pdf) of IONet: Learning to Cure the Curse of Drift in Inertial Odometry