Marcus Furuholmen

Evolution of Impulse Bursts Noise Filters

The paper deals with evolutionary design of impulse burst noise filters. As proposed filters utilize the filtering window of 5x5 pixels, the design method has to be able to manage 25 eight-bit inputs. The large number of inputs results in an evolutionary algorithm not able to produce reasonably working filters because of the so-called scalability problem of evolutionary circuit design. However, the filters are designed using an extended version of Cartesian Genetic Programming which enables to reduce the number of inputs by selecting the most important of them. Experimental evaluation of the method has shown that evolved filters exhibit better results than conventional solutions based on various median filters.

Seaeye Sabertooth A Hybrid AUV/ROV offshore system

Increasing use and complexity of subsea installations has put focus on the costs of maintaining these systems. In addition, access to these systems is sometimes limited by adverse weather and ice conditions. Conventional methods for intervention, maintenance and repair (IMR) using surface ships and ROVs are very expensive furthermore are response and mobilization times slow. To address this Saab Underwater Systems is in the process of developing a hovering Hybrid AUV/ROV system to remotely perform IMR without or strongly reduced need for a supporting ship. This system is based on the Double Eagle SAROV, a hovering Hybrid AUV/ROV in production for the military market and proven components from Saab Seaeye ROV product range.

Evolutionary approaches to the three-dimensional multi-pipe routing problem: a comparative study using direct encodings

In this study, three Genetic Algorithms (GAs) are applied to the Three-dimensional Multi-pipe Routing problem. A Standard GA, an Incremental GA, and a Coevolutionary GA are compared. Variable length pipelines are built by letting a virtual robot move in space according to evolved, fixed length command lines and allocate pipe segments along its route. A relative and an absolute encoding of the command lines are compared. Experiments on three proposed benchmark problems show that the GAs taking advantage of the natural problem decomposition; Coevolutionary GA, and Incremental GA outperform Standard GA, and that the relative encoding works better than the absolute encoding. The methods, the results, and the relevant parameter settings are discussed.

Scalability, generalization and coevolution -- experimental comparisons applied to automated facility layout planning

Several practical problems in industry are difficult to optimize, both in terms of scalability and representation. Heuristics designed by domain experts are frequently applied to such problems. However, designing optimized heuristics can be a non-trivial task. One such difficult problem is the Facility Layout Problem (FLP) which is concerned with the allocation of activities to space. This paper is concerned with the block layout problem, where the activities require a fixed size and shape (modules). This problem is commonly divided into two sub problems; one of creating an initial feasible layout and one of improving the layout by interchanging the location of activities. We investigate how to extract novel heuristics for the FLP by applying an approach called Cooperative Coevolutionary Gene Expression Programming (CCGEP). By taking advantage of the natural problem decomposition, one species evolves heuristics for pre-scheduling, and another for allocating the activities onto the plant. An experimental, comparative approach investigates various features of the CCGEP approach. The results show that the evolved heuristics converge to suboptimal solutions as the problem size grows. However, coevolution has a positive effect on optimization of single problem instances. Expensive fitness evaluations may be limited by evolving generalized heuristics applicable to unseen fitness cases of arbitrary sizes.

Coevolving heuristics for the Distributor's Pallet Packing Problem

Efficient heuristics are required for on-line optimization problems where search-based methods are unfeasible due to frequent dynamics in the environment. This is especially apparent when operating on combinatorial NP-complete problems involving a large number of items. However, designing new heuristics for these problems may be a difficult and time-consuming task even for domain experts. Therefore, automating this design process may benefit the industry when facing new and difficult optimization problems. The Distributors Pallet Packing Problem (DPPP) is the problem of loading a pallet of non-homogenous items coming off a production line and is an instance of a range of resource-constrained, NP-complete, scheduling problems that are highly relevant for practical tasks in the industry. Common heuristics for the DPPP typically decompose the problem into two sub-problems; one of pre-scheduling all items on the production line and one of packing the items on the pallet. In this paper we concentrate on a two dimensional version of the DPPP and the more realistic scenario of having knowledge about only a limited set of the items on the production line. This paper aims at demonstrating that such an unknown heuristic may be evolved by Gene Expression Programming and Cooperative Coevolution. By taking advantage of the natural problem decomposition, two species evolve heuristics for pre-scheduling and packing respectively. We also argue that the evolved heuristics form part of a developmental stage in the construction of the finished phenotype, that is, the loaded pallet.

Indirect Online Evolution --- A Conceptual Framework for Adaptation in Industrial Robotic Systems

A conceptual framework for online evolution in robotic systems called Indirect Online Evolution (IDOE) is presented. A model specie automatically infers models of a physical system and a parameter specie simultaneously optimizes the parameters of the inferred models according to a specified target behavior. Training vectors required for modelling are automatically provided online by the interplay between the two coevolving species and the physical system. At every generation, only the estimated fittest individual of the parameter specie is executed on the physical system, hence limiting both the evaluation time, the wear out and the potential hazards normally associated with direct online evolution (DOE), where every candidate solution has to be evaluated on the physical system. Features of IDOE are demonstrated by inferring models of a simple hidden system containing geometric shapes that are further optimized according to a target value. Simulated experiments indicate that the fitness of the IDOE approach is generally higher than the average fitness of DOE.

Continuous Adaptation in Robotic Systems by Indirect Online Evolution

A conceptual framework for online evolution in robotic systems called indirect online evolution (IDOE) is presented. A model specie automatically infers models of a hidden physical system by the use of gene expression programming (GEP). A parameter specie simultaneously optimizes the parameters of the inferred models according to a specified target vector. Training vectors required for modelling are automatically provided online by the interplay between the two coevolving species and the physical system. At every generation, only the estimated fittest individual of the parameter specie is executed on the physical system. This approach thus limits both the evaluation time, the wear out and the potential hazards normally associated with direct online evolution (DOE) where every individual has to be evaluated on the physical system. Additionally, the approach enables continuous system identification and adaptation during normal operation. Features of IDOE are illustrated by inferring models of a simplified, robotic arm, and further optimizing the parameters of the system according to a target position of the end effector. Simulated experiments indicate that the fitness of the IDOE approach is generally higher than the average fitness of DOE.

A Coevolutionary, Hyper Heuristic approach to the optimization of Three-dimensional Process Plant Layouts — A comparative study

A Coevolutionary, Hyper Heuristic approach to the optimization of Three-dimensional Process Plant Layouts (3DPPLs) is explored. By taking advantage of the natural problem decomposition, one population of layout heuristics, and another population of scheduling heuristics are coevolved. Generalized heuristics are evolved by training on multiple small problem instances, so that training time is reduced. The best generalized heuristic builds arbitrary sized 3DPPLs which reduce the cost by 18% when compared to a handmade heuristic. Specialized heuristics are evolved by optimizing each problem instance and outperforms the generalized heuristics after a fixed number of generations. Compared to a direct-encoded Genetic Algorithm, the benefit of specialized heuristics increases with the size of the problem, and costs are reduced by 30% when compared to the handmade heuristic.

An indirect approach to the three-dimensional multi-pipe routing problem

This paper explores an indirect approach to the Three- dimensional Multi-pipe Routing problem. Variable length pipelines are built by letting a virtual robot called a turtle navigate through space, leaving pipe segments along its route. The turtle senses its environment and acts in accordance with commands received from heuristics currently under evaluation. The heuristics are evolved by a Gene Expression Programming based Learning Classifier System. The suggested approach is compared to earlier studies using a direct encoding, where command lines were evolved directly by genetic algorithms. Heuristics generating higher quality pipelines are evolved by fewer generations compared to the direct approach, however the evaluation time is longer and the search space is more complex. The best evolved heuristic is short and simple, builds modular solutions, exhibits some degree of generalization and demonstrates good scalability on test cases similar to the training case.