The research article proposes a solution to integrate Visual Semantic Navigation (VSN) models into real robots. The authors have released a ROS-based framework for VSN, ROS4VSN, which can be used to deploy any VSN model in a ROS-compatible robot. They have tested this framework with two different robots and found a noticeable performance difference when tested in real-world and simulation environments. The ultimate aim is to improve the performance and efficiency of embodied agents in real-world scenarios.
Publication date: 29 Nov 2023
Project Page: https://github.com/gramuah/ros4vsn
Paper: https://arxiv.org/pdf/2311.16623
