Week 18 Overview

During this eighteenth week, I completed my final tasks related to the RADI engine of the Unibotics platform within JdeRobot.

Firstly, regarding the optimization of Nvidia GPU acceleration, a development team member discovered an issue in RADI where executing any command using the GPU resulted in an error indicating that the device was invalid, preventing proper GPU usage. Initially, I tested this on my Windows computer where the commands worked fine. My preliminary conclusion was that the issue might be due to the team member’s computer having both an integrated Intel GPU and an Nvidia GPU, possibly causing conflicts when using two GPUs.

In a second test on my Linux computer, I encountered the same error described by the team member. I tried launching RADI without the --device: /dev/dri variable, and it started functioning as it did on my Windows machine. This led me to conclude that the issue might be specific to Linux or that the use of this variable is not properly implemented, causing problems. I communicated all of this to the development team for their consideration.

Additionally, I was tasked with creating tables that contain performance and hardware consumption values for each exercise within RADI. These tables are to be included in the user documentation, providing reference values for users depending on their hardware configurations.

Therefore, I created three different tables: one with the characteristics of my Windows computer running RADI in CPU-only mode, another with my Windows computer running RADI in CPU + GPU mode, and a third with my Linux computer running RADI in CPU-only mode. I then incorporated these tables into the user documentation.

Lastly, I want to express my gratitude to my mentor and the development team for teaching me so much and making me feel comfortable throughout my internship at JdeRobot.