Terrain evaluation using CNN to generate a discretized matrix of probabilities with zones related to stability and potential support points for legged robots. Both terrain (gravel, dirt road, etc.) and obstacles are characterized probabilistically (0 - 100% probability of overcoming) generating a gradient in their surrounding areas. This information serves as a basis for making decisions regarding the selection of support points for legged robots.
Processed and original terrain:
