Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


36905.pdf (3.67 MB)

ETD Abstract Container

Abstract Header

Robot Localization Using Artificial Neural Network Under Intermittent Positional Signal

Saxena, Anujj
http://rave.ohiolink.edu/etdc/view?acc_num=ucin1613731846524738

Abstract Details

2020, MS, University of Cincinnati, Engineering and Applied Science: Mechanical Engineering.
Unmanned Aerial Vehicles (UAV) are gaining attention in the civilian domain with their numerous potential applications. This has been demonstrated recently in light of developments around the pandemic, where UAVs were used by law enforcement departments of various countries of the world. Multinational Corporations such as Mercedes Benz partnered with Matternet for drone-based deliveries in Switzerland. Ford recently filed a patent for a drone system that can be integrated with a car that could provide emergency services. UAVs rely very much on positional signals for navigation. Positional signals such as a global positioning system (GPS) are susceptible to an outage for periods ranging from one second to a minute. This work provides a novel approach by introducing an Artificial Neural Network (ANN) in the cases where there are long gaps in positional signal received by a UAV. During our prior research, similar problems were manifesting during bridge inspection during flights flown by the drones. Even in our experiments with indoor localization systems using `Decawave’, we faced similar problems. Decawave comprises Ultra-Wide-band modules that use Positioning and Networking Stack (PANS), a software library, that implements the Two-Way-Ranging method for localization. In the proposed work, an ANN is trained on drone dynamics for a pre-traveled path. Then this pre-trained network, during flight, uses back-propagation to update its weights/parameters in an online fashion, where-by it learns to “fill in” the GPS signal gaps by predicting the dynamics. In the event of a GPS Signal loss, this ANN, receiving the current state of the body as input, performs a forward propagation to predict the rigid body dynamics for the next state. The online learning capability ensures that this ANN’s weights are updated to reflect changing dynamics arising from changes such as different payloads. The results highlight a comparative study between a drone that implements only Extended Kalman Filter (EKF) and one that uses the suggested new approach with ANNs, and show the advantages of the proposed approach over the traditional EKF based approaches.
Manish Kumar, Ph.D. (Committee Chair)
Janet Jiaxiang Dong, Ph.D. (Committee Member)
David Thompson, Ph.D. (Committee Member)
84 p.
Mechanical Engineering
Unmanned Aerial Vehicles; GPS; Artificial Neural Network; Positional signal; Back-propagation; online learning

Recommended Citations

Citations

  • Saxena, A. (2020). Robot Localization Using Artificial Neural Network Under Intermittent Positional Signal [Master's thesis, University of Cincinnati]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1613731846524738

    APA Style (7th edition)

  • Saxena, Anujj. Robot Localization Using Artificial Neural Network Under Intermittent Positional Signal. 2020. University of Cincinnati, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=ucin1613731846524738.

    MLA Style (8th edition)

  • Saxena, Anujj. "Robot Localization Using Artificial Neural Network Under Intermittent Positional Signal." Master's thesis, University of Cincinnati, 2020. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1613731846524738

    Chicago Manual of Style (17th edition)

ucin1613731846524738
246
© 2020, all rights reserved.
This open access ETD is published by University of Cincinnati and OhioLINK.