E-VAC

E-VAC or you won't be back. E-VAC is an image processing convolutional neural network designed for raspberry pi in order to identify and map potential hazards during the event of building collapse.

Comment

What is E-VAC?

E-VAC was conceived with the endeavor of utilizing modern path planning and machine learning methods to mitigate the risk for first responders engaging with hazardous environments.

E-VAC consists of two major parts: a simulated robot exploring a path through a potentially dangerous environment, and a convolution neural network model which analyzes the risk of said path through classification of images taken by the robot in the course of its exploration.

How Does it Work?

Through the use of MATLAB ROS and navigation toolboxes in conjunction with Webots, we were able to successfully model an environment closely resembling that of the partially collapsed, inflamed, or otherwise dangerous buildings first responders have to face in real life. A rapidly expanding random tree model is then used to find a path through this environment, and pictures taken by the robot at key points in the path are analyzed by a convolutional neural network to identify and classify fire and environmental hazards.

Challenges we ran into

The construction of the environment for E-VAC was particularly strenuous as we had to acclimate to a new software, Webots. Through experimentation and tenacity, we were able to effectively design a model environment using this software for simulation.

What's next for E-VAC

In order to improve upon our model accuracy, we are working on compiling a more extensive database to train our neural network. Implementation with raspberry pi is also on the horizon as we plan to have it successfully navigate real world scenarios.

Additionally, we endeavor to shorten the branches of the rapidly expanding random tree model in order to create a more efficient navigation algorithm.

Built With

Share this project:

Updates