chammaryscanlime · 8:26 · 8 years ago
The Tuco Flyer flies, but it cannot yet fly itself. Its obstacle sensors sit unused, its winches fight over tension, and the camera swings like a pendulum after each move. Low flight stays a human job, because automatic mode cannot yet keep the cables clear of furniture.
Pendulum stability — The center of mass sits below the cable attachment point because the camera hangs half a meter below the sensors on a carbon fiber boom. The robot therefore swings and takes time to stop wobbling after each move. The designer accepted this trade-off to fly lower in the room, and the hardware is modular, so a counterweight or lower attachment point remains available if software cannot damp the motion.
Winch tension balancing — The current position estimate assumes a perfect cube with the flyer at its center, and the asymmetric pulley anchors make the error worse toward the edges. One winch repeatedly hits its lower tension limit. The slow correction lags when it is too gentle and causes oscillation when it is too fast.
Obstacle sensing — The flyer carries IR proximity sensors, laser rangefinders, an orientation sensor, and an accelerometer, all currently ignored. Automatic flight near the ground is limited because the sensors and cord attachments at the saucer must stay above every obstacle.
Gimbal tracking — The first tracker used a proportional loop that was sensitive to noise and produced dull framing. The current version runs multiple PID loops per axis, triggered when the region of interest crosses gain zones at the screen edges. The gimbal still jerks when the region of interest is large relative to the frame.