The first Robot Group "Computer-Blimp", demonstrated at Robofest I (Fall 1988, Austin TX), consisted of an airframe made from aluminum and plastics suspended from an envelope made of polyester survival blankets. A small onboard computer directed the activity of the machine by controlling the speed and direction of two propellers as well as the pitch of the motors. Thrust vectors were selected at random as the blimp wandered aimlessly throughout the exhibit area.

Also demonstrated at Robofest I was the "Fish-Blimp". While the Computer Blimp brandished the independence of computer control, the Fish Blimp provided the visual impact of a flapping fin for its propulsion, as well as that of a sleek envelope design. Steering was achieved with a transverse propeller mounted at the nose. The machine was controlled manually through an umbilical cord, which provided power as well.

The next year and a half yielded significant progress for the blimps. The computer blimp, now dubbed the Mark II, was given a new envelope constructed of helium- proof balloon material and much more powerful motors, with larger propellers. Its computer system was modified to take commands from a remote terminal through an RS-232 umbilical cord, while a scanning ultrasonic range-finder was used to collect data about the environment. The command stream and range data were combined to provide training data for a neural network, the planned autonomous control mechanism. The collected data, however, proved to be useless due to the weight of the umbilical interfering with true free-flight behavior.

Along with the Mark II, the "Bipedal Ornithopter" made its debut at the "Air and Space Expo/Cyberspace Convention" (Spring 1990, Austin TX). The Bipedal Ornithopter used its 'running legs' for launching, and its "flapping wings" for propulsion. The ship carried its own power and was controlled by radio. It was capable of very elaborate maneuvering, which enhanced the already extreme visual impact of the design.

By 1991, the Computer Blimp project, now called the Mark III, had evolved further, with the design of a custom computer system for control, radio-based training exercises, improved thruster design, extra ultrasonic sensors, and a much larger envelope, now measuring three meters long by one meter in diameter. The Robot Group had acquired access to a building with a large atrium, consisting of four story heights, arches large enough to fly through, and several bridges: a perfect training ground. Many hours of training data were collected over a period of months, then finally fed to a back-propagation neural network training program. The results were less than were hoped. While evidence of the neural net's reaction to new data was present, it was insufficient to handle the job of navigating the blimp. The Mark III was demonstrated at Robofest II in February, 1991 and at Motorola, in Austin, later that spring.

Since that time, the Mark IV improvements have been made. The size of the envelope has been increased (6 meters by 1.5 meters) to provide the lift necessary to carry the CCD camera, video transmitter and extra power required for telepresence. Several new sensor scanning patterns for "netability" have been designed but have not yet been tested, since access to the building used for training has been lost.

Current projects include a strictly radio-controlled "tractor" blimp, with a robot arm, for rescue missions, and a "Geodesic-Sphere" airship, with internal thrust mechanisms, capable of acrobatic maneuverability. Investigations into laser range-finder vision, air-flow measurement, and the use of a flux-gate compass, for the purpose of providing better data to the neural networks, are ongoing.

The track record of the Robotic Airship Project demonstrates a clear path of progress toward the defined goal. The unique propulsion systems and envelope designs provide a visual impact which implies a life-form, while the navigation problems have pushed the Robot Group into state-of-the-art research of control and behavior technologies.