Project Overview   Subsystems   Test Flights   CSSS presentation
Abstract: Skeyeball is a platform for undergraduate and masters level projects in embedded systems. It is an airborne camera platform for short range video imaging. Various objectives include basic infrastructure, microflyer reconnaisaance, automomous flight, automated stability control, image stabilization, vision, etc.
Skeyeball is also a viable platform for research experimentation, but that is not its primary purpose at this time.
Supervisor: Steven D. Johnson (email@example.com) Scientist: Danko Antolovic (firstname.lastname@example.org)
Caleb Hess (email@example.com)
Current Participants: Past participants: Prashant Patel (summer 2000) Aaron Wilson (summer 2001) Credit: available by arrangement at all levels; team projects are encouraged. Background: For all projects, the design team needs background in some or all of the following areas: digital design, FPGA implementation, system-level programming, event handling software, embedded software, network communications. There are projects that require domain-specific knowledged (such as vision), and other projects that are simply system design challenges. Engineering support for the Skeyeball us limited, so project so project proposal evaluations will consider the support needs, and may result in suggested refinements to the project goals.
A human observer (OBS) wants to investigate a nearby, inaccessible location whose view is blocked by the terrain. OBS launches the airplane and flies it by remote control to the area of interest, using direct sight. Once the airplane reaches the vicinity of the target, the video telemetry is used by the ground pilot. During this phase, autonomous flight control is used to keep the camera platform stable in flight; and OBS reconnoiters by directing the camera to look around. Once an interesting ground object is sighted, OBS directs the on-board system to track it with the camera while OBS resumes flight control. OBS may also order the aircraft to perform a vision based task, such as orbiting the target.Later extensions of the project may provide the means to autonomously search and identify objects, but the present goal is simpler: to use the tracking subsystem to guide the airplane. Similarly tasks may become more complex. For instance, the airplane may purposely or accidentally fly out of line-of-sight control, and have on-board capability to climb into line-of-sight, or even return to base.
The subsystems described below are potential projects already identified. However Skeyeball is intended to be a platform for individual projects in system design. The project is open to anyone who would find it an appropriate vehicle for research investigations of any kind.
Tracking behavior was clearly observable on some of these straight-line passes.
|Video clips (.mov format)|
The primary goal was to attempt using a longer range (F??) camera lens to enable higher-altitude passes over the target, in order to allow longer tests of the tracking system. No meaningful data was recorded for two reasons:
Danko was in Bloomington for two weeks and made some revisions to the visions system control. The main enhancement was ground control of the camera gimbal. It was believed that the resulting ability to orient the camera would make it easier to initiate tracking and result in a higher percentage of useful passes in flight. Given the very limited development time, the enhancement was made primarily in software with minor hardware modifications enabling additional radio channels to be interfaced to the Coldfire processor.
June 1. Conditions: fair. In three flights*, several passes were made over the ground target using the long range camera lense, with the intent of allowing higher altitude circling. At higher altitudes, the circling radius could be larger and banking angles reduced, so piloting consequently would be much easier. However, with the much narrowed field of view, it was nearly impossible for the vision system operator to locate the target. When the target was found, it was visible for only a fraction of a second, not enough reaction time for the operator to put the system into tracking mode. Use of the long-range camera lense was abandoned.
*Power was taken from the vehicle battery. The battery charge was exhaused prematurely because the engine cooling fan was turned on in "accessory" mode.
June 2 Conditions: Marginal; overcast, with gusty winds from SW at 5-20 kph. Several flights were done, each with multiple passes. The lower range lense was used on the camera. In the first flights circling was done at three altitudes, and the best seemed to be at about 40 m [Bryce?].
On the second flight, all attempts at circling were at the intermediate altitude. The operator was able to find the target and had time to activate the tracking system. However, radio interference, glare, and lack of contrast on the monitor made this task difficult. In none of the circling passes was evidence of tracking behavior observed. There was an unexplained jump in camera position when tracking was activated, and sometimes the activation command failed to take effect. Danko felt that target size and contrast were adequate for target identification.
In the final three flights, repeated linear passes were made over the target. Passes into the wind were quite good from the perspective of experimentation, since ground speed was extremely low. However, none of the passes in which tracking was activated showed clear evidence of tracking behavior. Recorded evidence was less conclusive than in prior test flights because the ground operator could move the camera as well as the tracking system.