AVANT - UFMG

AVANT - UFMG is a team at the Federal University of Minas Gerais composed by students from varying backgrounds with the focus on doing research about autonomous aerial vehicles. The team builds and program drones to do autonomous tasks, participating in competitions in the field. Further information about the team can be found on their website (portuguese).

Aerial vehicles made by the group. On the left, flying-wings. Middle, quadricopter. Right, fixed-wing.

I joined the group in my first semester and spent three years there. I was initially allocated to the electronic/software development division. My main responsability involved path-planning and navigation, especifically to develop fast and autonomous three-dimensional path-planning algorithms for drones navigation, using sensing information such as GPS and evading fixed and dynamic obstacles. For that purpuse I adjusted Dijkstra, and after, A* algorithms to suit the operations restrictions. Being part of the electronic/software division I was also responsible for doing the electronic assembly, and was able to learn about computer vision, specifically how the group would use it to detect shapes and streets to land. For street detection was also tested Machine Learning techniques.

On the left, explains how grid is created around vehicle for path-planning. Middle, color and shape detection using computer vision. Right, a good street detection using computer vision.

During my final year in the team, I led the group as a Manager. My roles included managing the project and specially the team and the members. During this time, because the members were primarily among our classmates and within the engineering field, I started to strongly advertise the group across various campus sectors. This resulted in applications from unfamiliar students and from diverse backgrounds such as mathematics, administration, and even arts. Concurrently, I personally created the first website and two “institutional” videos. Additionally, I facilitate the creation of technical reports and share our results, making them publicly available for other researchers, for example here. After this joint effort in making the team more recognizable in our university, we were finally invited, among others famous groups, to present the group for all engineering freshers.

On the left, the drone doing the task of navigating through the tube and detecting the colors. Right, scheme of the arena, by ROBOCUP.

My final contribution was securing the participation and participating for the first time in the Brazilian Robotics Competition 2020: Petrobras Robotic Challenge/Robocup Brazil. Due to COVID-19, the competition was conducted in simulation. The objective was to enable a drone to autonomously perform multiple tasks in an arena, including inspecting oil tubes, landing on platforms, and scanning and reading information from them. There, I was mainly responsible for navigation, utilizing sensor and computer vision data to map the environment, ensuring safe navigation in the arena, precise alignment and accurate landing in designated spots. It was employed ROS and GAZEBO.