Location

Room # EB402

Start Date

5-6-2019 9:10 AM

Description

The ability to fly a remote control (RC) aircraft from a third-person perspective is a skill that many hobbyists and enthusiasts enjoy. With a little practice, an RC pilot can sense the aircraft’s orientation and apply the correct inputs for it to orbit, achieve specific mission objectives, or line-of-sight waypoints. The work done in [1] proved that third-person sensing of an aircraft’s attitude is possible. This work seeks to improve the deep learning methods, increase hardware capabilities through the addition of a turret and motion capture validation, and add control algorithms for complete autonomous flight. While the results are not comprehensive, this report gives the work accomplished thus far. Overall this report will show that substantial progress has been made towards third-person autonomous control.

Comments

Session 1

Share

COinS
 
May 6th, 9:10 AM

Third-Person Autonomous Control using Deep RC

Room # EB402

The ability to fly a remote control (RC) aircraft from a third-person perspective is a skill that many hobbyists and enthusiasts enjoy. With a little practice, an RC pilot can sense the aircraft’s orientation and apply the correct inputs for it to orbit, achieve specific mission objectives, or line-of-sight waypoints. The work done in [1] proved that third-person sensing of an aircraft’s attitude is possible. This work seeks to improve the deep learning methods, increase hardware capabilities through the addition of a turret and motion capture validation, and add control algorithms for complete autonomous flight. While the results are not comprehensive, this report gives the work accomplished thus far. Overall this report will show that substantial progress has been made towards third-person autonomous control.