Discipline: Computer Sciences and Information Management
Subcategory: Nanoscience
Darius Wills-Browne - Hampton University
In nanotechnology, due to size and difficulty of its fine design, the trend is to have a systematic self-assembled methodology that can design the structure of the nano-scale robots. We adopt the behavior of nature, specifically the natural evolution for the purpose of self-sustained heuristics that can be effectively behaved in the environment. In the area of computational intelligence, evolutionary algorithm (EA) is a technique that mimics the genetics behavior to solve real world problems. We use the EA technique to mimic biological gene mutation and gene selection through an initial set of potential solutions, and gradually evolve them into the ‘fittest’ solution for the problem. In here, we have used Hill Climber technique. Evolutionary Robotics uses EA to find the fittest robot in a given population to survive. Next, it chooses the fittest robots and disposes of ones with a lower fitness value. The robots left are modified and improved to be better in the next generation/ iteration, similar to how offspring are an improved version of a parent on fitting the environment. The process repeats until goal is reached. Hill Climber algorithm helps on altering robots with low fitness to ones with higher fitness to fit with the environment. Nanobots can have different specific tasks. In every generation, robot agents will be evaluated to assure the surviving ones can better achieve that specific task. Start with a specific vector filled with random numbers between 0 and 1 associated with the genes of one generation. Then, the fitness will be computed as some representative of the vector. The next generation will be created by simply employing a small amount of mutation with the chance of 5% or less. In the selection process from current generation of mutated parent robots, the fittest ones will be selected to be the next generation’s offspring robots.
Using Hill Climber technique, 50 parent robots are randomly selected, then mutated, and the best are selected by comparison of their fitness values to generate offspring robots. The process is repeated 5000 generations. We monitor the collective behavior of robots in each generation, along with the dynamic of fitness values. The graphical results demonstrates that the fitness value reaches above 95% of its highest peak value within the first 200 generations, then gradually approach the highest peak value. Using a grayscale graph, the behavior of all robots are monitored to check its dynamic over 5000 generations. The results demonstrate the direction of fitness is towards having the fittest condition to the environmental. In conclusion, using self-assembling bio-inspired evolutionary techniques, we can convert the design process of nanobots autonomous and non-supervised reducing the hassle of nanoscale fine design, and adopt the natural behavior of nature by mimicking biological evolution. Using Hill Climber evolutionary technique, it is shown that the use of mutation and selection operator would make the design process efficient and autonomous. As a future research question and further research, we need to adopt applicable real-world problems – such as a nanobot that targets specifics molecular structure like cancerous-patterned cells – as the objective function for evaluating the fitness function and test the bio-inspired selfassembling heuristic in the real world application problem.
Funder Acknowledgement(s): Dr. Michelle Claville, Dr. Jean Muhammad, and Dr. Moayed Daneshyari all from Hampton University. This research was funded by NanoHU: Nanoscience Transforming STEM Education at Hampton University, funded by National Science Foundation HBCU-UP ACE Implementation Award HRD – 1238838.
Faculty Advisor: Moayed Daniel Daneshyari,