The Hawaii Space Flight Laboratory had a problem. For the Leonidas low-cost nano-satellite program, the University of Hawaii program needed to design an adapter to connect as many as 18 small satellites to larger satellite. Upon attaining the proper orbital or launch location, the different satellites would emerge and go into their own orbits. The HSFL design team had come up with a structure that fit the space constraints of the project but weighed 22 kilograms. The specification for the structure was no more than 20 kilograms.
The design team tapped Marcelo Kobayashi, an assistant professor or mechanical engineering at the University of Hawaii at Manoa and an expert in high performance computing and advanced mathematics. Kobayashi had been working on software packages that could take a design and feed the specifications into a new type of genetic algorithm that mimicked the processes of evolution at the cellular development level. The packages are called BioTOM, short for biologically inspired Topography Optimization Method. (Those who attended the recent TechHui Summit saw more about his work).
The software optimizes designs in a new way that the HSFL team hoped would solve their weighty problem. When Kobayashi's programs had completed their work, the resulting adapter design closely resembled the previous one and provided similar structural integrity. But the optimized design only weighed 9 kilos, a remarkable 41 percent reduction. In the world of small satellites, such tremendous weight savings would save considerable launch fuel or allow researcher to place one or two additional nanosatellites on the primary satellite vehicle. "Each pound costs tens of thousands of dollars to put in space. We saved hundreds of thousands of dollars in mass reduction," explains Kobayashi.
To achieve this astonishing result, Kobayashi didn't need a supercomputer. The optimization was performed with an eight-processor desktop used for workstation and standard graphics processing tasks. BioTOM may herald a new era of design optimization that relies on a whole new way of looking at genetic algorithms and natural evolutionary processes. "It goes beyond simple biomimetics. In that, you look at an existing design and try to mimic that design. You can look at a bird, see how they fly and imiate that design in airplanes. I am trying to go one step further. Instead of looking at and copying from existing designs I am trying to understand the evolutionary mechanisms that led to that design in the first place," says Kobayashi.
How is this different? Traditional use of genetic algorithms in design has relied more on inputting parameters around a design from nature, setting the software motion, and looking for a single optimal result by iterating over and over again on the same process. BioTOM, in contrast, maps the parameters of a design down to a quasi-cellular level, mimicking structures of wings, capillaries, or other natural forms down to fine-grained detail. In other words, BioTOM looks at each design problem as topology and seeks to create the best support structure to create a functional topology or an alernate topology that fulfills the same goals. In a more visual sense, BioTOM optimizes a design in similar fashion to the way nature evolves and selects for complex structures in wings, leaves, and circulatory systems.
The program then will take those core topological elements and run through multiple iterations to examine alternatives and then build upon each result to improve on the initial topology map. Thus, the process kicks natural selection into high gear with the help of silicon and smart logic. "If you look at any organism, they do not pop up in the end result," says Kobayashi. "You don't emerge from the womb as an adult. You need cellular divisions and developmental stages. Those developmental stages are controlled by the genetics encoded in our DNA. I interpret DNA as a program but as a set of rules that govern a cellular division process."
Ultimately, Kobayashi hopes BioTOM will signal the start of a paradigm shift in the design of complex systems as engineers of all strips can tap into the power that Mother Nature has to date solely wielded and democratize evolution into a desktop commodity available to all. Kobayashi hopes to being tweaking BioTOM to enable parallel processing on computers.