Integrating Continuous Differential Evolution with Discrete Local Search for Meander Line RFID Antenna Design

Result sets produced in the application of Differential Evolution to the problem of meander line RFID antenna design, as reported in the in prep. paper "Integrating Continuous Differential Evolution with Discrete Local Search for Meander Line RFID Antenna Design". Two files of generated solutions are included: the eight local search variants applied to the 7x7 instance (7x7_ls_study.txt.zip); and all other results for variants of DE applied to 5x5 to 10x10 problems, plus the results of the prior ACO (de_w_ls_for_RFID_results.txt.zip). Results are grouped by the algorithm composition used to generated them. Also included is C++ source code for a version of the NEC++ antenna evaluator, a Linux executable for the NEC evaluator (mynec), and a Python script for transforming node paths into input files for NEC++ (evaluate.py).

The two data files are tab-delimited with the following columns:
alg: the algorithm combination used
size: grid size (5-10)
trial: for DE this is the random seed, for ACO this is the value of the q0 parameter
f0: solution's resonant frequency
e: solution's efficiency (%)
length: number of nodes in the antenna path
path: node path as a space-delimited list

Values for alg include:
DE: the multiobjective DE 'control'
DE + bias: DE with solution archive selection bias
DE + bias + ls x/y/n: DE with bias and backbite local search, using the x/y/n solution reintegration strategy (see paper for details)
DE + bias + (uncounted) ls, regen/det/3: DE with best-performing local search strategy run under similar conditions to prior ACO
ACO: the prior Ant Colony System algorithm; due to the way solutions were represented in that work some duplicate solutions were produced

The Python evaluation script can be used interactively to generate the input used by NEC or, at the command line, it will also attempt to run the NEC simulator and report the objective values for the given solution.