Evolvable Systems for Space Applications

This article surveys the research of the Evolvable Systems Group at NASA Ames Research Center. Over the past few years, our group has developed the ability to use evolutionary algorithms in a variety of NASA applications ranging from spacecraft antenna design, fault tolerance for programmable logic chips, atomic force field parameter fitting, analog circuit design, and earth observing satellite scheduling. In some of these applications, evolutionary algorithms match or improve on human performance. 1 Spacecraft Antenna Design In this section we summarize a proof-of-concept study [12] that investigated the optmization of the deployed antenna on the Mars Odyssey spacecraft. Automated antenna synthesis via evolutionary design has recently garnered much attention in the research literature [13]. Evolutionary algorithms show promise because, among search algorithms, they are able to effectively search large, unknown design spaces. NASA’s Mars Odyssey spacecraft is currentlyin Martian orbit. Onboard the spacecraft is a quadrifilar helical antenna that provides telecommunications in the UHF band with landed assets, such as robotic rovers. This antenna can be seen in Fig. 1. Each helix is driven by the same signal which is phase-delayed in 90◦ increments. A small ground plane is provided at the base. It is designed to operate in the frequency band of 400438 MHz. Based on encouraging previous results in automated antenna design using evolutionary search, we wanted to see whether such techniques could improve upon Mars Odyssey antenna design. Specifically, a coevolutionary genetic algorithm is applied to optimize the gain and size of the quadrifilar helical antenna. Figure 1: Photograph of the quadrifilar helical UHF antenna deployed on the Mars Odyssey spacecraft. The optimization was performed in-situ – in the presence of a neighboring spacecraft structure [9]. On the spacecraft, a large aluminum fuel tank is adjacent to the antenna. Since this fuel tank can dramatically affect the antenna’s performance, we leave it to the evolutionary process to see if it can exploit the fuel tank’s properties advantageously. Optimizing in the presence of surrounding structures would be quite difficult for human antenna designers, and thus the actual antenna was designed for free space (with a small ground plane). In fact, when flying on the spacecraft, surrounding structures that are moveable (e.g., solar panels) may be moved during the mission in order to improve the antenna’s performance. 1.1 Experiments and Results Experiments were set up as follows. The Numerical Electromagnetics Code program was used to evaluate all antenna designs. We used a parallel master/slave generational genetic algorithm with a population size of 6000. One point crossover across byte boundaries was used at a rate of 80%. Mutation was uniform across bytes at a rate of 1%. Runs were executed on 32-node and 64-node Beowulf computing clusters. The wire geometry encoded by each individual chromosome was first translated into a NEC input deck, which was subsequently sent to the NEC simulator. The segment size for all elements was fixed at 0.1λ, where λ was the wavelength corresponding to 235 MHz. A coarse model of the neighboring fuel tank was used in the simulations. Its size and position was calculated based on engineering drawings of the spacecraft. To compare our results to the spacecraft antenna, we modeled that antenna with the best data we had at the time of this writing. A coevolutionary genetic algorithm was applied to the quadrifilar helical antenna optimization. Two populations are used: one consisting of antenna designs, and one consisting of target vectors. The fundamental idea is that the target vectors encapsulate levelof-difficulty. Then, under the control of the genetic algorithm, the target vectors evolve from easy to difficult based on the level of proficiency of the antenna population. Each target vector consists of a set of objectives that must be met in order for a target vector to be “solved.” A target vector consisting of two values: the average gain (in dB), VSWR, and antenna volume. A target vector was considered to be solved by a given antenna if the antenna exceeds the performance thresholds of all target. Values for target gain ranged between -50 dB (easy) and 8 dB (difficult). Target VSWR values ranged between 100 (easy) and 20 (difficult). Target antenna volumes ranged from 100,000 cm (easy) to 100 cm (difficult). Target vectors are represented as a list of floating point values that are mutated individually by randomly adding or subtracting a small amount (5% of the largest legal value). Single point crossover was used, and crossover points were chosen between the values. Antennas are rewarded for solving difficult target vectors. The most difficult target vector is defined to be the target vector that only one antenna can solve. Such a target vector garners the highest fitness score. Target vectors that are unsolvable, or are very easy to solve by the current antenna population, are given low fitness scores. Fitness was expressed as a cost function to be minimized. The calculation was as follows: F = −GL + ∑ (C ∗ Vi) C = { 0.1 if Vi ≤ 3 1 if Vi > 3 where: GL = lowest gain of all frequencies measured at θ = 0◦ and φ = 0◦, Vi = VSWR at the ith frequency. Lacking from this calculation was a term involving sidelobe/backlobe attenuation. We chose not include such a term because we reasoned that as the mainlobe gain increased, the sidelobes/backlobes would decrease in size. A set of five runs were executed using the algorithm described above. Only one of the runs found an antenna design that exceed that benchmark antenna. Fig. 2 shows the gain plots for both the evolved and actual Mars UHF antennas. Fig. 3 show the antennas, structures, and radiation patterns of actual Mars Odyssey UHF and evolved antenna. The evolved antenna measures 6cm × 6cm × 16cm which approximately four times as small volumewise as the benchmark (roughly 10cm × 10cm × 25cm). At 400 MHz, the average gain of the evolved antenna was 3.77 dB and 1.95 for the benchmark antenna. At 438 MHz, the average gain of the evolved antenna was 2.82 dB and 1.90 for the benchmark antenna. This represent a 93% improvement at 400 MHz and a 48% improvement at 438 MHz in the average gain.