Benjamin E. Beckmann

Evolving coordinated quadruped gaits with the HyperNEAT generative encoding

Legged robots show promise for complex mobility tasks, such as navigating rough terrain, but the design of their control software is both challenging and laborious. Traditional evolutionary algorithms can produce these controllers, but require manual decomposition or other problem simplification because conventionally-used direct encodings have trouble taking advantage of a problems regularities and symmetries. Such active intervention is time consuming, limits the range of potential solutions, and requires the user to possess a deep understanding of the problems structure. This paper demonstrates that HyperNEAT, a new and promising generative encoding for evolving neural networks, can evolve quadruped gaits without an engineer manually decomposing the problem. Analyses suggest that HyperNEAT is successful because it employs a generative encoding that can more easily reuse phenotypic modules. It is also one of the first neuroevolutionary algorithms that exploits a problems geometric symmetries, which may aid its performance. We compare HyperNEAT to FT-NEAT, a direct encoding control, and find that HyperNEAT is able to evolve impressive quadruped gaits and vastly outperforms FT-NEAT. Comparative analyses reveal that HyperNEAT individuals are more holistically affected by genetic operators, resulting in better leg coordination. Overall, the results suggest that HyperNEAT is a powerful algorithm for evolving control systems for complex, yet regular, devices, such as robots.

Harnessing Digital Evolution

In digital evolution, self-replicating computer programs-digital organisms-experience mutations and selective pressures, potentially producing computational systems that, like natural organisms, adapt to their environment and protect themselves from threats. Such organisms can help guide the design of computer software.

Investigating whether hyperNEAT produces modular neural networks

HyperNEAT represents a class of neuroevolutionary algorithms that captures some of the power of natural development with a computationally efficient high-level abstraction of development. This class of algorithms is intended to provide many of the desirable properties produced in biological phenotypes by natural developmental processes, such as regularity, modularity and hierarchy. While it has been previously shown that HyperNEAT produces regular artificial neural network (ANN) phenotypes, in this paper we investigated the open question of whether HyperNEAT can produce modular ANNs. We conducted such research on problems where modularity should be beneficial, and found that HyperNEAT failed to generate modular ANNs. We then imposed modularity on HyperNEATs phenotypes and its performance improved, demonstrating that modularity increases performance on this problem. We next tested two techniques to encourage modularity in HyperNEAT, but did not observe an increase in either modularity or performance. Finally, we conducted tests on a simpler problem that requires modularity and found that HyperNEAT was able to rapidly produce modular solutions that solved the problem. We therefore present the first documented case of HyperNEAT producing a modular phenotype, but our inability to encourage modularity on harder problems where modularity would have been beneficial suggests that more work is needed to increase the likelihood that HyperNEAT and similar algorithms produce modular ANNs in response to challenging, decomposable problems.

HybrID: a hybridization of indirect and direct encodings for evolutionary computation

Evolutionary algorithms typically use direct encodings, where each element of the phenotype is specified independently in the genotype. Because direct encodings have difficulty evolving modular and symmetric phenotypes, some researchers use indirect encodings, wherein one genomic element can influence multiple parts of a phenotype. We have previously shown that Hyper-NEAT, an indirect encoding, outperforms FT-NEAT, a direct-encoding control, on many problems, especially as the regularity of the problem increases. However, HyperNEAT is no panacea; it had difficulty accounting for irregularities in problems. In this paper, we propose a new algorithm, a Hybridized Indirect and Direct encoding (HybrID), which discovers the regularity of a problem with an indirect encoding and accounts for irregularities via a direct encoding. In three different problem domains, HybrID outperforms HyperNEAT in most situations, with performance improvements as large as 40%. Our work suggests that hybridizing indirect and direct encodings can be an effective way to improve the performance of evolutionary algorithms.

Evolution of an adaptive sleep response in digital organisms

Adaptive responses to resource availability are common in natural systems. In this paper we explore one possible evolutionary cause of adaptive sleep/wake behavior. We subjected populations of digital organisms to an environment with a slowly diminishing resource and recorded their ability to adapt to the changing environment using sleep. We also quantified the selective pressure not to sleep in this competitive environment.We observed that diminishing resource availability can promote adaptive sleep responses in digital organisms even when there is an opportunity cost associated with sleeping.

Evolution of Cooperative Information Gathering in Self-Replicating Digital Organisms

We describe a study in the application of digital evolution to the problem of cooperative information gathering. In digital evolution, self-replicating computer programs evolve to perform tasks and optimize resource usage in order to survive within a user defined computational environment. Instruction-level mutations during replication and CPU-cycle rewards for desired behavior produce natural selection within the population. The evolution is open-ended and not limited by human preconceptions. The contributions of this work are (1) to demonstrate that cooperative information gathering can evolve in digital organisms and (2) to provide insight into the fundamental processes governing evolution of such behavior.

Automatically generating adaptive logic to balance non-functional tradeoffs during reconfiguration

Increasingly, high-assurance software systems apply self-reconfiguration in order to satisfy changing functional and non-functional requirements. Most self-reconfiguration approaches identify a target system configuration to provide the desired system behavior, then apply a series of reconfiguration instructions to reach the desired target configuration. Collectively, these reconfiguration instructions define an adaptation path. Although multiple satisfying adaptation paths may exist, most self-reconfiguration approaches select adaptation paths based on a single criterion, such as minimizing reconfiguration cost. However, different adaptation paths may represent tradeoffs between reconfiguration costs and other criteria, such as performance and reliability. This paper introduces an evolutionary computation-based approach to automatically evolve adaptation paths that safely transition an executing system from its current configuration to its desired target configuration, while balancing tradeoffs between functional and non-functional requirements. The proposed approach can be applied both at design time to generate suites of adaptation paths, as well as at run time to evolve safe adaptation paths to handle changing system and environmental conditions. We demonstrate the effectiveness of this approach by applying it to the dynamic reconfiguration of a collection of remote data mirrors, with the goal of minimizing reconfiguration costs while maximizing reconfiguration performance and reliability.

Evolution of resistance to quorum quenching in digital organisms

Quorum sensing (QS) is a collective behavior whereby actions of individuals depend on the density of the surrounding population. Bacteria use QS to trigger secretion of digestive enzymes, formation and destruction of biofilms, and, in the case of pathogenic organisms, expression of virulence factors that cause disease. Investigations of mechanisms that prevent or disrupt QS, referred to as quorum quenching, are of interest because they provide a new alternative to antibiotics for treating bacterial infections. Traditional antibiotics either kill bacteria or inhibit their growth, producing selective pressures that promote resistant strains. In contrast, quorum quenching and other so-called anti-infective strategies focus on altering behavior. In this article we evolve QS in populations of digital organisms, a type of self-replicating computer program, and investigate the effects of quorum quenching on these populations. Specifically, we injected the populations with mutant organisms that were impaired in selected ways to disrupt the QS process. The experimental results indicate that the rate at which these mutants are introduced into a population influences both the evolvability of QS and the persistence of an existing QS behavior. Surprisingly, we also observed resistance to quorum quenching. Effectively, populations evolved resistance by reaching quorum at lower cell densities than did the parent strain. Moreover, the level of resistance was highest when the rate of mutant introduction increased over time. These results show that digital organisms can serve as a model to study the evolution and disruption of QS, potentially informing wet-lab studies aimed at identifying targets for anti-infective development.

Evolving cooperative pheromone usage in digital organisms

The use of chemicals to communicate among organisms has enabled countless species, from microorganisms, to colonies of insects, to mammals, to survive and flourish in their respective environments. Ants, arguably natures most successful exploiters of this behavior, have evolved the use of pheromones to communicate in a wide range of situations, including mating, colony recognition, territory marking, and recruitment to new nest sites and food sources. We examine the evolution of the use of pheromones to aid in the location of, and migration to, a target area by groups of digital organisms. In an initial set of experiments, these organisms evolved efficient patterns of exploration that obviated the need for pheromones. When evolved in a more adverse environment, organisms again evolved effective search strategies, but also evolved the use of pheromones to enable the task to be completed by group members more quickly and with fewer movements. We also show that evolved organisms are more robust and better able to react to a change in the environment than a handbuilt solution. This work demonstrates the complexities that exist in the evolution of pheromone-enabled cooperation and provides insight into the behaviors executed by seemingly simple organisms in nature.

Evolution of Adaptive Population Control in Multi-agent Systems

Dynamic population management is an important aspect of multi-agent systems. In artificial immune systems, for example, a shortage of agents can lead to undetected threats, while an overabundance of agents can degrade quality of service and if unchecked, even create new vulnerabilities. Unfortunately, designing an effective strategy for population management is complicated by the myriad of possible circumstances and environmental conditions the agents may face after deployment. In this paper, we present the results of a study in applying digital evolution to the population management problem. In digital evolution, populations of self-replicating computer programs evolve in a user-defined computational environment, where they are subject to mutations and natural selection. Our results demonstrate that populations of digital organisms are capable of evolving self-adaptive replication behaviors that respond to attack fluctuations, as well as clever strategies for cooperating to mitigate attacks. This study provides evidence that digital evolution may be a useful tool in the design of self-organizing and self-adaptive agent-based systems.