Monday, June 17, 2013

Anti-Jamming Beamforming Arrays Optimized with Genetic Algorithms: A Recap

Today, I've decided to give a recap on the research that I do since I've written on several subjects during the last week. I'm researching different stochastic algorithms for optimizing anti-jamming beamforming arrays. To date, I've researched the effectiveness of genetic algorithms at doing this task. I've included a YouTube video that I created a while back that explains how the array works.

The above array consists of four antennas, phase shifters, step attenuators, a 4-way RF power combiner, and digital controllers. The array operates by adjusting phase and attenuation on the three outer antennas. The end goal is to focus RF energy on the signal of interest while simultaneously minimizing (i.e., nulling) RF energy in jammer directions. I used a genetic algorithm (GA) to optimize my anti-jamming antenna array. The genetic algorithm (GA) controls the phase shifter and step attenuator settings based on survival of the fittest mate selection, bit-wise chromosomal crossover, and bit-wise chromosomal mutations. The figure below shows a flowchart of the GA. In my research application, the GA uses signal to interference and noise ratio (SINR) as its fitness function.

Figure 1: Flowchart showing how the GA operates to maximize SINR (Copyright Jonathan Becker)
In the figure above, the signal of interest that you want to receive is the signal portion of SINR, and the jammers are the Interference portion. An initial population of binary settings are created, decoded into hardware settings, and evaluated to create the SINR fitness functions. Crossovers and mutations are applied with probabilities p(cross) and p(mutation). I assume that noise is additive white Gaussian noise (SINR) and is dependent on the system bandwidth and ambient temperature. The new solutions (i.e., children) are decoded into hardware settings and reevaluated. The whole process is repeated until a solution with an acceptable fitness function is found or the algorithm reaches a maximum number of iterations (i.e., generations). If you wish to learn more about how the GA works, please read my previous post on genetic algorithms.

In future research, I will compare other stochastic algorithms such as Particle Swarm Optimization (PSO) with the GA. I will likely build a new antenna array build with surface mount components on printed circuit boards (PCBs). This array will have eight antennas to increase the maximum number of jammers it can thwart. I'm also considering using wideband antennas, so the array can operate on multiple bands. Come back and visit my blog, as I plan on writing on these and other technical subjects, and I will give other tips on surviving graduate school.

Your truly,

Jonathan Becker
ECE PhD Candidate
Carnegie Mellon University