Skip to Main Content
 

Global Search Box

 
 
 

ETD Abstract Container

Abstract Header

A NOVEL FRAMEWORK TO EFFICIENT PATH PLANNING THROUGH REAL-TIME COST MAP GENERATION USING NEURAL NETWORKS FOR SEARCH AND RESCUE MISSIONS

Machina, Keith Martin Kinyua
http://orcid.org/0000-0002-4281-7532
http://rave.ohiolink.edu/etdc/view?acc_num=kent1716813727184678

Abstract Details

2024, MS, Kent State University, College of Arts and Sciences / Department of Computer Science.
Search and rescue missions are critical endeavours aimed at locating survivors trapped in the aftermath of various calamities such as earthquakes, landslides, tsunamis etc. The time efficiency required for these missions is pivotal in saving lives, prompting the adoption of robotics with advanced technology to expedite operations. Effective planning and strategizing play key roles in enhancing a robot’s efficiency, particularly in prioritizing areas of need. Typically, search and rescue missions often present a high level of complexity due to a multitude number of factors to be considered when planning and optimizing operations. Factors such as determining hot-spot regions, harsh weather conditions, terrain complexity, environmental hazards, and time sensitivity, just to a mention few, introduce intricacies to the task of effective planning. For instance, a factor such as environmental conditions is considered dynamic as it may change from time to time. Additionally, the criticality of different regions may change dynamically as soon as the information is made available, further complicating the task for both rescue robots and even fast responders. This research addresses the challenge of optimizing path planning in search and rescue missions, treating it as a variant of the travelling salesman problem and proposes a hybrid technique that incorporates both algorithmic and non-algorithmic techniques, to tackle the problem. The hybrid technique leverages a tweaked version of the U-net neural network that is trained on two pieces of information: static data, data that encircles topological map data such as slope and dynamic data deduced from map findings data that would be vital in singling out hotspots and priority regions on the map. Blending these two pieces of information, an amalgamated cost, a value that incorporates both priority and cost of traversal, is determined to aid robot path planning decisions. Training a neural network on this data enables it to predict the associated amalgamated cost map, permitting real-time path planning through a real-time amalgamated cost map that is generated during the robot’s mission.
Jong-Hoon Kim (Advisor)
Hassan Raiful (Committee Member)
Gorkona Sharma (Committee Member)
77 p.
Artificial Intelligence; Robotics
cost map, amalgamated cost map, reinforcement learning

Recommended Citations

Citations

  • Machina, K. M. K. (2024). A NOVEL FRAMEWORK TO EFFICIENT PATH PLANNING THROUGH REAL-TIME COST MAP GENERATION USING NEURAL NETWORKS FOR SEARCH AND RESCUE MISSIONS [Master's thesis, Kent State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=kent1716813727184678

    APA Style (7th edition)

  • Machina, Keith. A NOVEL FRAMEWORK TO EFFICIENT PATH PLANNING THROUGH REAL-TIME COST MAP GENERATION USING NEURAL NETWORKS FOR SEARCH AND RESCUE MISSIONS. 2024. Kent State University, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=kent1716813727184678.

    MLA Style (8th edition)

  • Machina, Keith. "A NOVEL FRAMEWORK TO EFFICIENT PATH PLANNING THROUGH REAL-TIME COST MAP GENERATION USING NEURAL NETWORKS FOR SEARCH AND RESCUE MISSIONS." Master's thesis, Kent State University, 2024. http://rave.ohiolink.edu/etdc/view?acc_num=kent1716813727184678

    Chicago Manual of Style (17th edition)

kent1716813727184678
77
© 2024, some rights reserved.Creative Commons License A NOVEL FRAMEWORK TO EFFICIENT PATH PLANNING THROUGH REAL-TIME COST MAP GENERATION USING NEURAL NETWORKS FOR SEARCH AND RESCUE MISSIONS by Keith Martin Kinyua Machina is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License. Based on a work at etd.ohiolink.edu.
This open access ETD is published by Kent State University and OhioLINK.
Release 3.2.12