Control of height and speed using visual sensing in a UAV
In past work, non-visual sensing modes (e.g. GPS) have been combined with visual sensing to determine either range given speed or speed given range. The problem of how vision might be used in a standalone mode without a secondary sensing mode has not been tackled. For example, is it possible to use image motion to determine height above terrain when there is no available measure of ground speed? Preliminary experimental work suggests that bees deliberately impart a lateral sinusoidal displacement of known amplitude upon their flight path in order to generate image motion in a defined way. As the velocities corresponding to the lateral wiggle are a function of the amplitude, the bees would be able to deduce range from the lateral image motion and thereby control their height. Apparently, bees can also regulate their flight speed based on the longitudinal image motion. With range known from lateral optic flow and forward speed known from longitudinal optic flow, the bees are able to regulate both speed and height using the one sensory system. The objective of this project is to investigate whether such as scheme could be implemented on an unmanned helicopter.
Description of Work:
- Using existing in-house rotorcraft simulation, develop a robust scheme for controlling the height and speed of a helicopter from simulated image motion.
- Refine the algorithm to incorporate obstacle avoidance techniques. Implement and test on a UNSW Canberra unmanned helicopter.