Monocular Visual Odometry for Assistive Lower Limb Exoskeletons

Assistive Lower Limb Exoskeletons (LLEs) offer significant potential for individuals with impaired lower limb mobility. However, current LLEs are limited by their reliance on predefined movements that do not account for the surrounding environment. In our previous solution we employed an {RGB-D} camera to develop an environment-adaptive gait planning system that only considered the latest detection to reason upon the environment, which is unsuitable in many scenarios. In this work, we aim to upgrade our environment understanding system by employing deep learning methods for monocular visual odometry (VO): in this way, the environment can be mapped, greatly expanding gait planning capabilities, while also lowering the camera requirements.