Скачать или смотреть Underground Mine Semantic Segmentation for Robotic Path Planning

This is an example of applied deep learning, specifically semantic segmentation, in the underground mining environment for autonomous robot navigation.

Here the drivable path is estimated to assist autonomous vehicles to plan the path that they should take in the mine.

There are two examples provided, one is in a main decline that has lighting from a light vehicle and then a more challenging scenario where a robot is travelling through a conveyor drive in a return air way. The environment is very dusty and the lighting is not ideal.

The inference from the neural networks feed into occupancy maps of the environment to provide pixelwise classification for objects for use in probabilistic path planning modules.