Terrain characterization is a vital skill for mobile robots as it enables them to plan their paths, navigate complex terrains, optimize their locomotion, conserve energy, interact with the environment, and ensure safety. By understanding the terrain, robots can adapt and perform more effectively in a wide range of applications.
A wide range of traditional techniques has been employed to classify and characterize terrains for use with mobile robots, but most are based on the terrains properties directly, and not on how they affect the robot. This thesis aims to gather sensor data on a real robot and use it as a basis to build a machine learning model on how the terrain affects the robot. Different directions can be taken with the work depending on what the student is most interested in.
