Couceiro M., Ghamisi P. Fractional Order Darwinian Particle Swarm Optimization. Applications and Evaluation

Springer, 2016. — 82 p.Biomimetics has been exploited in several research areas as a means to endow artificial systems with intelligence, resilience, adaptation, and natural selection typically exhibited by living organisms and biological ecosystems, in order to solve complex human problems through new technologies inspired by such biological systems at macro-and nanoscales.
Bioinspiration has received special attention in the robotics community for the past two decades in order to solve complex optimization problems through bioinspired algorithms. It has been mainly devoted to the study of robot swarms comprising many unsophisticated robots interacting locally with neighbor robots and the environment, which can exhibit useful collective patterns resembling the way swarms of biological species behave collectively to strive for survival in hostile environments against threats and predators. However, such techniques have a broader application range, not being confined only to swarm robotics.
One of the most well-known bioinspired optimization techniques is particle swarm optimization (PSO), which has demonstrated remarkably high potential in optimization problems wherein conventional optimization techniques cannot find a satisfactory solution, due to nonlinearities and discontinuities. The PSO technique consists of a number of particles whose collective dynamics, resembling a biological ecosystem, allows effectively exploring the search space to find the optimal solution. The Darwinian PSO (DPSO) is an evolutionary optimization algorithm and an extension of the original PSO that makes use of Darwin’s theory of natural selection to regulate the evolution of the particles and of their collective dynamics, so that complex optimization of functions exhibiting many local MAXIMA/minima can be successfully accomplished. The fractional order DPSO (FODPSO) incorporates in DPSO the notion of fractional-order derivatives to attain memory of past decisions and even better convergence properties.Particle Swarm Optimization
Fractional-Order Darwinian PSO
Case Study I: Curve Fitting
Case Study II: Image Segmentation
Case Study III: Swarm Robotics
Conclusions