Pilar Caamaño

Autonomous UAV based search operations using Constrained Sampling Evolutionary Algorithms

This paper introduces and studies the application of Constrained Sampling Evolutionary Algorithms in the framework of an UAV based search and rescue scenario. These algorithms have been developed as a way to harness the power of Evolutionary Algorithms (EA) when operating in complex, noisy, multimodal optimization problems and transfer the advantages of their approach to real time real world problems that can be transformed into search and optimization challenges. These types of problems are denoted as Constrained Sampling problems and are characterized by the fact that the physical limitations of reality do not allow for an instantaneous determination of the fitness of the points present in the population that must be evolved. A general approach to address these problems is presented and a particular implementation using Differential Evolution as an example of CS-EA is created and evaluated using teams of UAVs in search and rescue missions. The results are compared to those of a Swarm Intelligence based strategy in the same type of problem as this approach has been widely used within the UAV path planning field in different variants by many authors.

Real-valued multimodal fitness landscape characterization for evolution

This paper deals with the characterization of the fitness landscape of multimodal functions and how it can be used to choose the most appropriate evolutionary algorithm for a given problem. An algorithm that obtains a general description of real valued multimodal fitness landscapes in terms of the relative number of optima, their sparseness, the size of their attraction basins and the evolution of this size when moving away from the global optimum is presented and used to characterize a set of well-known multimodal benchmark functions. To illustrate the relevance of the information obtained and its relationship to the performance of evolutionary algorithms over different fitness landscapes, two evolutionary algorithms, Differential Evolution and Covariance Matrix Adaptation, are compared over the same benchmark set showing their behavior depending on the multimodal features of each landscape.

JEAF: A Java Evolutionary Algorithm Framework

There are not many tools in the evolutionary computing field that allow researchers to implement, modify or compare different algorithms. Additionally, those tools usually lack flexibility, maintenance or some other characteristic, so researchers program their own solutions most of the time, reimplementing algorithms that have already been implemented hundreds of times. This paper introduces a new framework for evolutionary computation called JEAF (Java Evolutionary Algorithm Framework) that tries to offer a platform to facilitate the tasks of comparing, analyzing, modifying and implementing evolutionary algorithms, reusing components and programming as few as possible. JEAF also aims to be a tool for evolutionary algorithm users that employ these algorithms to solve other problems not related with evolutionary computation. In this sense, JEAF provides methods to distribute an evolutionary process and to plug external tools to perform the evaluation of candidate solutions.

Specialization analysis of embodied evolution for robotic collective tasks

Abstract The objective of this work is to analyze embodied evolution based algorithms in coordinated multi-robot tasks that require specialization. This type of algorithm performs a Darwinian open-ended evolution where the individuals that make up the population are embodied in the physical robots and situated in an environment. The robots interact autonomously in an asynchronous fashion, leading to a complex dynamic system in continuous evolution with dependencies among parameters that make theoretical studies of specialization quite difficult in real cases. Consequently, the aim here is to perform a theoretical analysis of this type of embodied evolution based algorithms, establishing a set of canonical parameters that define their operation. A generic algorithm of this type is designed that allows us to formally study the relevance of the canonical parameters. In this paper this study concentrates on specialization for the construction of heterogeneous robotic teams. The conclusions obtained in the theoretical framework are confirmed in a real multi-robot collective gathering task using one of the many real embodied evolution based algorithms and showing that two canonical parameters are the most relevant in terms of specialization for this type of algorithms. Some insights into how to adjust these canonical parameters in a real problem are provided.

A Complex Systems Based Tool for Collective Robot Behavior Emergence and Analysis

This paper presents a Complex Systems Theory based methodology and tool for the automatic design of multiagent or multirobot collective behaviors for the optimized execution of a given task. The main goal of this methodology is the representation of a generic task to be optimally performed in a Complex Systems simulator called WASPBED and the subsequent analysis of the emergent states thus obtained. This way, by tweaking environmental parameters in the system, the behaviors of the different collective behaviors obtained can be studied. The example used to test the methodology deals with collective behaviors for optimized routing in unknown environments.

Augmenting the NEAT algorithm to improve its temporal processing capabilities

This paper is concerned with the incorporation of new time processing capacities to the Neuroevolution of Augmenting Topologies (NEAT) algorithm. This algorithm is quite popular within the robotics community for the production of trained neural networks without having to determine a priori their size and topology. However, and even though the algorithm can address temporal processing issues through its capacity of establishing feedback synaptic connections, that is, through recurrences, there are still instances where more precise time processing may go beyond its limits. In order to address these cases, in this paper we describe a new implementation of the NEAT algorithm where trainable synaptic time delays are incorporated into its toolbox. This approach is shown to improve the behavior of neural networks obtained using NEAT in many instances. Here, we provide some of these results using a series of typical complex time processing tasks related to chaotic time series modeling and consider an example of the integration of this new approach within a robotic cognitive architecture.

A Behavior Based Architecture with Auction-Based Task Assignment for Multi-robot Industrial Applications

The study of collective robotic systems and how the interaction of the units that make them up can be harnessed to perform useful tasks is one of the main research topics in autonomous robotics. Inspiration for solutions in this realm can be sought in nature and in the interaction of natural social systems whether through simple trading strategies or through more complex economic models. Here we present a three level behavior based architecture for the implementation of multi-robot based cooperation systems that is based on the individual, the collective and the social levels. In particular, here we are going to consider the application of this architecture for the implementation and study of auction-based strategies for assigning tasks in a real application of multi-robot systems. Our approach is more focused on studying the behavior of auction-based techniques from an engineering point of view in terms of parameters and results analysis. To this end, we have used a real industrial case as an experimental platform where a heterogeneous group of robots must clean a ship tank. The results obtained show how the performance of the auction mechanism we have implemented does not degrade in terms of computational cost when the number of robots is increased, and how the complexity of the task assignment can be highly increased without any change in the cooperative control system.

Solving a heterogeneous fleet vehicle routing problem with time windows through the asynchronous situated Coevolution algorithm

In this work we present the practical application of the Asynchronous Situated Coevolution (ASiCo) algorithm to a special type of vehicle routing problem, the heterogeneous fleet vehicle routing problem with time windows (HVRPTW). It consists in simultaneously determining the composition and the routing of a fleet of heterogeneous vehicles in order to serve a set of time-constrained delivery demands. The ASiCo algorithm performs a situated coevolution process inspired on those typical of the Artificial Life field that has been improved with a strategy to guide the evolution towards a design objective. This strategy is based on the principled evaluation function selection for evolving coordinated multirobot systems developed by Agogino and Tumer. ASiCo has been designed to solve dynamic, distributed and combinatorial optimization problems in a completely decentralized way, resulting in an alternative approach to be applied to several engineering optimization domains where current algorithms perform unsatisfactorily.

Scalable Task Assignment for Heterogeneous Multi-Robot Teams

This work deals with the development of a dynamic task assignment strategy for heterogeneous multi-robot teams in typical real world scenarios. The strategy must be efficiently scalable to support problems of increasing complexity with minimum designer intervention. To this end, we have selected a very simple auction-based strategy, which has been implemented and analysed in a multi-robot cleaning problem that requires strong coordination and dynamic complex subtask organization. We will show that the selection of a simple auction strategy provides a linear computational cost increase with the number of robots that make up the team and allows the solving of highly complex assignment problems in dynamic conditions by means of a hierarchical sub-auction policy. To coordinate and control the team, a layered behaviour-based architecture has been applied that allows the reusing of the auction-based strategy to achieve different coordination levels.

Task-driven species in evolutionary robotic teams

This paper deals with the problem of obtaining coordinated behavior in multirobot systems by evolution. More specifically, we are interested in using a method that allows the emergence of different species if they are required by the task, that is, if specialization provides an advantage in the completion of the task, without the designer having to predefine the best way to solve it. To this end, in this work we have applied a co-evolutionary algorithm called ASiCo (Asynchronous Situated Co-evolution) which is based on an open-ended evolution of the robots in their environment. In this environment the robots are born, mate and die throughout the generations as in an artificial life system. In order to show that ASiCo is capable of obtaining species automatically if they are advantageous, here we apply it to a collective gathering and construction task where homogeneous teams are suboptimal.