From Heightmaps to Cameras: Teacher-Student Reinforcement Learning for Rover Navigation: Master’s thesis
Translated title
From Heightmaps to Cameras: Teacher-Student Reinforcement Learning for Rover Navigation
Author
Term
4. semester
Education
Publication year
2025
Submitted on
2025-06-04
Pages
69
Abstract
This thesis investigates a DAgger-based teacher-student framework for transferring navigation behavior from a policy trained on privileged heightmap input to one using noisy RGB-D observations. The goal is to support learning under realistic sensor condi- tions, where reinforcement learning on RGB- D data remains challenging due to its partial and high-dimensional nature. A simulation pipeline was extended to enable separate sen- sor inputs for teacher and student using the RobuROC4 platform. While the student in- creasingly aligned with the teacher’s actions, it failed to generalize or perform the task effectively. This is attributed to memory- bound dataset handling and limited rollout diversity. The findings suggest that teacher- student imitation learning holds promise for sensor modality transfer, but depends on scalable infrastructure capable of supporting large and diverse training datasets.
Keywords
Reinforcement learning ; Imitation learning ; teacher-student ; DAgger ; Isaac Sim ; Isaac Lab ; RLRoverLab ; RobuROC4 ; RB Summit ; Leo Rover ; Simulation ; Camera ; RGB-D ; Heightmap
Documents