Ph.D. Proposal: Caleb Harris

Thu Dec 09 2021 02:30 PM
Collaborative Visualization Environment (CoVE)
"Risk-Aware Planning and Obstacle Avoidance for Aerial Systems During Emergency Response Scenarios"

You're invited to attend

 

Ph.D. Proposal

 

Caleb Harris

(Advisor: Prof. Dimitri Mavris)

 

"Risk-Aware Planning and Obstacle Avoidance for Aerial Systems During Emergency Response Scenarios"

 

Thursday, December 9
2:30 p.m.

Collaborative Visualization Environment (CoVE)
Weber Space Science and Technology Building (SST II)

 

Abstract:
Over the last decade, natural disasters and climate-related disasters have caused the deaths of over 400,000 people internationally and around one trillion dollars of damage in the US alone. Researchers predict these numbers to rise in the future from the impacts of climate change and urbanization. There is a demand for a more effective and efficient disaster response to decrease the loss of life and mitigate the cost. Autonomous robots and unmanned aerial systems have successfully provided support and situational awareness during disasters in the past thirty years. However, recent disasters have shown that they must be capable of rapid monitoring over large areas and in challenging environments to remove first responders from dangerous situations successfully. Therefore, recent work has sought to employ multiple systems that can autonomously navigate complex and dynamic environments while considering the reliability and cost for first responders and stakeholders. This work analyzes three phases of autonomous flight: path planning, obstacle avoidance, and coordination. Networked aerial, autonomous systems, or UAS swarms, are applied for disaster response situational awareness and search missions by leveraging offboard data and onboard sensing and applying a multi-phase approach using global and local planning. This work addresses these two phases independently to analyze the best algorithms and data sources for this application.

The first hypothesis is that the global planning stage can provide preflight flight plans that improve safety by maximizing publicly available data through deep learning and fusion algorithms and optimizing offline planning through risk-aware and multi-agent algorithms. The second hypothesis is that data-driven, vision-based avoidance algorithms trained and tested in simulated environments can produce high-speed flight performance for low-cost systems during onboard trajectory planning between preflight waypoints. Previous work has investigated these techniques for applications like emergency landing, autonomous racing, and more, but little work has directly explored the disaster response scenario, particularly in simulated environments. This work performs simulated experiments that focus on vision-based navigation, risk-aware planning, and multi-agent coordination. A set of scenarios evaluate preflight and inflight planning performance while including environment and model uncertainty. The hope is that this work demonstrates a successful implementation of this framework and soon can reduce cost and increase lives saved during disaster response scenarios across the globe.

 

Committee:

  • Prof. Dimitri Mavris – School of Aerospace Engineering (advisor)
  • Prof. Daniel Schrage – School of Aerospace Engineering
  • Prof. Polo Chau – School of Computational Science and Engineering
  • Prof. Richard Vuduc – School of Computational Science and Engineering
  • Dr. Alexia Payan – School of Aerospace Engineering
  • Dr. Youngjun Choi – United Parcel Service (UPS), Inc.

Location

Collaborative Visualization Environment (CoVE)