Student: Guangyi Liu
Project: Classification-Aware Path Planning of Network of Robots
View: Research Poster (PDF) | Presentation (YouTube)
Department: Mechanical Engineering and Mechanics
Advisor: Nader Motee
Abstract
We propose a classification-aware path planning architecture that employs layers of convolutional neural network (CNN) and long short-term memory (LSTM) in a distributed fashion among a team of robots to solve a classification task using localized (partial) observations from the environment. In our proposed method, the learned neural network layers memorize the history of collected features by each agent and steer robots to visit the most informative regions. This will significantly reduce the total energy cost (e.g., battery life) when solving the classification problem. Moreover, a parallel design of the path planning structure reduces the training complexity significantly. Our findings have been validated with a map classification problem in the simulation environment of satellite campus maps using quadcopters with on-board cameras.
About Guangyi Liu
Guangyi Liu and Arash Amini are graduate students working with Prof. Nader Motee in the Department of Mechanical Engineering and Mechanics, Lehigh University. Martin Takac is currently an Associate Professor in the Department of Industrial and Systems Engineering at Lehigh University. Nader Motee is Professor in the Department of Mechanical Engineering and Mechanics, Lehigh University. Dr. Motee’s current research area is distributed control systems and real-time robot perception.