Airways: Optimization-Based Planning of Quadrotor Trajectories according to High-Level User Goals
summary
An interactive design tool turns sketched keyframes into physically feasible quadrotor flights by discretizing time, using a 4-DOF linear mass-point model and solving a quadratic program (add inequality bounds for accel/force); camera objectives make the problem nonconvex and are solved iteratively. It models a gimbal with an angular penalty and a perspective-cost, is validated indoors with a motion capture system, executes real flights (one shot rotates ~100 degrees while keeping the target in view), and fails on highly aggressive or moving-target scenes.
AI-written summary · about
chapters (10)
- 0:00Introduction to quadrotor filming challenges
- 1:34Design tool for flight planning
- 3:22Physical feasibility of trajectories
- 6:10Linear model for optimization
- 9:01Generating feasible trajectories
- 10:51Including camera constraints
- 12:18Perspective and aesthetic goals
- 13:43Real-world flight demonstrations
- 14:48Limitations and future work
- 15:29Conclusion and Q&A
related talks
57mReal-Time Decision Making in Hybrid Systems
17mDynamic Replanning of Multi-drone Missions using Dynamic Forward Slicing
09mWIP: Hijacking Attacks on UAV Follow-Me Systems in Realistic Scenarios
18mGridDrones: A Self-Levitating Physical Voxel Lattice for Interactive 3D Surface Deformations
46mPaparazzi, An Open Source UAV for Research
52mFlying services with a drone
48mFlying services with a drone
45m(Deep) Learning to Fly