July 24, 2007
Michael Fu described how labor intensive and time consuming repair work is in its current form. In 2004, there was a gas leak aboard the International Space Station that was 0.04 psi and took a 2-man crew of astronauts two weeks to locate. This is the motivation behind his and Edmond Wong’s research with cooperative robotics. The future exploration missions NASA is designing are becoming more long-term and reach farther out in space than we have already explored. This means that faster, more robust leak detection systems will be necessary. Mr. Fu and Mr. Wong propose that robots are a promising solution to this dilemma.
The current NASA robots in use are relatively large in size. Mr. Fu showed a couple examples of these robots: the Mars rovers and the Canada Arm. They are designed more for exploration than for repair work and so are very expensive and time-consuming to replace. The cost of the Canada Arm 1, for example, was $100 million. Future NASA robots are multi-functional single agents which have complex components, complex controls, and are also very expensive. They also require high energy and computing power. Two examples given were Robonaut and Spidernaut.
Inspiration for Cooperative Robotics
Microscopic organisms, insects, and other animals exhibit something referred to as swarm or flock intelligence/behavior. This is the inspiration behind the concept of cooperative robotics. Several examples of these organisms are: bees, leafcutter ants, ladybugs, herring that swim in schools, cranes migrating, and buffalo that stampede as a defensive mechanism. If a single bee attacked someone, it wouldn’t be as effective of an attack as if an entire swarm of bees attacked them. The same goes for the buffalo stampede. A stampede of several bulls is much more effective than just one bull attacking. This is the kind of behavior that they would like to harbor in their robotics.
The advantages to a system of this kind, using cooperative robotics is the increased level of redundancy, robustness, fault tolerance, simpler, smaller design, and scalability (if the size of the project increases then just send in more robots). The basic concept behind cooperative robotics is that there would be an increase in confidence behind inspection and repair work performed autonomously by teams of cooperative robots compared to the labor-intensive and time-consuming manual inspection that exists today.
Current Progress for Robotics Test Bed
At GRC, Dr. Wong and Michael work with cooperative robots in a test bed. They use an 8’ x 6’ arena and even have a simulated obstacle setup, with 3D real-time tracking. The little robots that they use are called Robo Pucks. They are six inches in diameter, use 2 x 8051 microprocessors, have a capacitance for proximity sensing, an optical sensor, a wireless transceiver, use Li-Po batteries, and have two independent motors. Each has a custom circuit board with an RF Transceiver with antenna. Michael passed around one of the Robo Pucks so that we could get a closer look.
Michael and Dr. Wong have implemented a communications protocol between the Robo Pucks. They have peer-to-peer, host-to-robot, and broadcasting forms of communication. They are very simple and robust. There is a modified version of Multiple-Access with Collision Avoidance (MACA), proposed by Karn in 1990. They use a Phasespace optical motion capture system with real-time tracking, LEDs mounted on robots to allow unique tracking, 480 frames per second, and 12.4 mega pixel resolution. Michael showed two videos representing their current progress with the robots. The first was regarding proximity sensing. In the video the white and yellow Robo Pucks were trying to avoid the green Robo Puck. The second video was a billiard simulation.
Dr. Wong and Michael plan to complete work soon regarding decreasing the size of the robots. They also want to explore adhesion technologies, research cooperative control algorithms and fault detection methods.
There was a brief question and answer session after the presentation, where several different topics were brought up:
There was a smaller generation of robots that had many of the same features. The problems that they faced with the smaller robots there were power issues being that they were so small. Also, the cooperative robots team is approaching algorithms for both a central hub that would control the robots and one that would have the robots work separately, similar to the case of natural animals/insects discussed previously. There are numerous, very broad applications in both the space and terrestrial environments for these cooperative robots. The last thing that Michael did was to show a demonstration of cooperative little cone robots searching for a target area using an algorithm. The little cone robots remembered where they had already explored and discovered certain targets very quickly, which is the design of the program.