Team project
Seoul, Korea
MATLAB University Student AI Competition: 3rd place (video)
We used MATLAB to generate path of excavation, hopefully automating the basic excavation in construction site
Process
With the average age of workers in construction sites increasing and numerous fatalities due to accidents, we recognized the urgent need for automation in construction site operations.
(1) Utilizing a Realsense depth camera, we captured the 3D surface information of the sand and extracted the necessary point cloud data for decision-making.
(2) We then converted this point cloud into a second-order curve to obtain parameters suitable for deep learning.
(3) Employing a neural network with three hidden layers using GRU (Gated Recurrent Unit), we successfully generated an excavation path replicating a human controller's expertise.
(4) We successfully integrated our developed excavation path model into the RC excavator.
My role involved designing the architecture of the neural network layers and conducting the training process using MATLAB.
Comments