*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
Jeffrey Pattison
(Advisor: Prof. Dimitri Mavris)
will propose a doctoral thesis entitled,
An Approach for Risk-Informed UAS Guidance in Urban Environments to Support First Responders
On
Thursday, June 2 at 2:00 p.m.
Collaborative Visualization Environment (CoVE)
Weber Space Science and Technology Building (SST II)
Abstract
In the past decade, Unmanned Aerial Systems (UAS) have been adopted to assist in a variety of tasks ranging from photography to agriculture. One area where the use of UAS is continually growing is in law enforcement, where they are used for search and rescue purposes, crime scene investigation, and first response. The benefits of using a UAS as a first responder include faster response times and the capability to relay vital information to ground personnel that they would not have access to otherwise. This not only improves police officer safety, but also has the potential to de-escalate situations and prevent police chases. Currently, the cities that use UAS as a first responder rely on trained UAS pilots to remain in communication with a teleoperator in the police department. The pilots are stationed throughout the city on building rooftops. Having at least one pilot per UAS is inefficient as the number of UAS increases, and more pilots require more coordination with the teleoperator. For efficient emergency response, the police departments would like completely autonomous UAS that do not rely on the pilot oversight. However, as the level of UAS autonomy increases and pilot intervention decreases, there is a greater need for UAS risk assessment and safety measures in the event of a failure. Therefore, the focus of this research includes approximating the risk of using a UAS to assist in route planning for an emergency response to create low-risk routes.
A common metric used for risk assessment in aviation is the probability of a fatality. Due to a lack of historical flight data for UAS, modeling and simulation methods are required to predict the probability of fatality for UAS. Many of the common methods found in literature rely on expensive physics-based models coupled with Monte Carlo simulations to estimate the risk. Relying on physics-based methods and Monte Carlo simulations becomes computationally expensive, making it difficult to use in real-time to respond to emergency events. The approach proposed explores using a Convolutional Neural Network with Multilayer Perceptron to create a surrogate model that can approximate the risk without requiring the physics-based models. This surrogate model can be used to assist in the route planning by quickly estimating the risk for given locations.
Aviation regulations stipulate aircraft operations cannot exceed an acceptable level of safety. With the UAS risk estimations, low-risk routes can be created to ensure an acceptable level of risk is not exceeded. The second objective is to explore several common route-planning algorithms, including A* and RRT*, to compare the expected safety levels of optimal routes created and to determine which method is most suitable for emergency use when applied to a large city.
The final objective of this work is to apply the previous efforts to a simulated system of UAS dispersed in a city to respond to emergency events. An auction-based task allocation method will be used with two different value metrics that determine which UAS will respond to a given task. The first metric will be based on distance to the task to allow the closest UAS to respond to an emergency while the second metric is expected level of risk, so the lowest risk UAS is deployed. The response time and estimated risk of the system of UAS will be compared when using the two different metrics in the task allocation. This research will not only provide a method to make risk-informed decision making for emergency response but will enable fast-time UAS risk approximation and route planning.
Committee
