After 2018 competition in the spring, the software, embedded, and airframe subteams have been hard at work improving upon our experience from the 2017–2018 school year. Here’s an update from each of our teams:

Software

With the removal of moving obstacles from this year’s competition rules, the software team can now reallocate energy to mission tasks such as object detection, localization, and classification as well as stationary obstacle avoidance.

For computer vision tasks, we are working on implementing and training different machine learning models to determine the model with the highest accuracy. Some of the models we are implementing are Mask R-CNN and YOLOv3 to crop images around the targets. To train and test our models, we are making and taking aerial images of our own targets to collect data, and holding data labeling meetings for use of the data with the models.

Triton UAS is a project team from the sunny campus of the University of California, San Diego. We are a student-run team that designs, builds, tests, and flies our own UAV to compete each year in the Student UAS Competition hosted by the AUVSI Seafarer Chapter against teams from around the world. Despite the fact that the 2017 competition coincided with finals week, our team placed 9th overall out of 42 teams.

After a year of hard work the UCSD AUVSI team went to Maryland with our backup plane to compete and to show off the work we accomplished during the school year. The competition pits 30 schools from around the world During the mission demonstration we were able to fly and complete image reconnaissance. Even when losing GPS lock during flight we were able to continue flying and complete the mission. The flight was a good learning experience on what we need to do in the future to be able to adapt to different mission configurations such as very large obstacles.