A Mission-Critical Communications Planning Over Contested RF Spectrum with Deep Reinforcement Learning Aided Artificial Intelligence

Overview

University of New Mexico Mission-critical communications (MCC) refer to those that support operations involving high risk to human life and property. As RF spectrum becomes highly contested, ensuring mission-success with MCC requires intelligent planning policies. This project develops a novel game-theoretic model for MCC and a Deep Q-Network (DQN) implemented Deep Reinforcement Learning (DRL) based Mission-Critical Communications Protocol (MCCP) to learn to complete a mission within given resource-constraints against an adversary. An example critical mission is defined as two radios exchanging messages within a given time constraint over two oppositely-directed communication links in the presence of an adversarial jammer. Mission-planning requires the radios to learn when and how to switch directions vs. channels based on the behavior of an adversary. Through extensive-form sequential game modeling, the problem was found too complex to solve analytically and beyond traditional reinforcement-learning due to uncountable state-space. DQN-implemented DRL is shown to be an ideal approach to learn effective policies for such MCC planning. Results on an actual wireless network showed that the DQN-implemented DRL could achieve mission-success with 0.9 probability. A new DRL algorithm called Deep Policy Hill Climbing was developed that outperformed the original DQN-DRL algorithm by 30%. Developed abstract game- theoretic model applies to a wide-range of critical mission scenarios including crop-planning and resource- allocation in management. Hence, the developed DRL mission planning framework can be utilized for ensuring mission-success in a wide-variety of fields, far beyond MCC. This makes the case for adapting DQN-powered AI techniques to mitigate the risk of human errors in high stake missions. Using MobileNet and Long-Short Term Memory Neural Networks to Help Correct Speaking Mouth Poses of Children with Cleft Palate Xingyu Ji Portsmouth Abbey School Portsmouth, Rhode Island Children with cleft palate often suffer from speech disorders. Children need guidance from professional speech therapists to teach them to pronounce correctly. Speech therapists and similar organizations are hard to find in many regions of the world. Even with the opportunity to receive surgeries, kids may still not be able to change their way of speaking. I realize the difficulties in completely altering the process of speech therapy, but I can use technology in the field of deep learning to increase the efficiency of it. The aim of my project is to construct a neural network that takes inputs from the patients, determine whether the input matches the correct mouth poses or not, and finally gives the patient feedbacks on each word. One of the most important aspects of speech therapy is to train kids to pose their mouths correctly while speaking. Even though the main process remains personal training with therapists, the duration of training can be shortened if children can practice on their own and gain immediate feedback on their performances. The main approach of this project is using convolutional neural networks and recurrent neural networks to process the data inputs, mostly video clips of patients speaking, and output to the patient if the input corresponds with correct poses that the model is trained on.

Competition history

• JSHS 2020 Category not listed

Resources

• National JSHS abstract book ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.