Wojciech Turek

Massively concurrent agent-based evolutionary computing

The fusion of the multi-agent paradigm with evolutionary computation yielded promising results in many optimization problems. Evolutionary multi-agent system (EMAS) are more similar to biological evolution than classical evolutionary algorithms. However, technological limitations prevented the use of fully asynchronous agents in previous EMAS implementations. In this paper we present a new algorithm for agent-based evolutionary computations. The individuals are represented as fully autonomous and asynchronous agents. An efficient implementation of this algorithm was possible through the use of modern technologies based on functional languages (namely Erlang and Scala), which natively support lightweight processes and asynchronous communication. Our experiments show that such an asynchronous approach is both faster and more efficient in solving common optimization problems.

Software Agent Systems for Improving Performance of Multi-Robot Groups

Agent paradigm is becoming more and more popular in design of complex multi-robot systems. Similarities between features attributed to robots and agents lead to defining an agent as a single robot control program or even calling a robot an “embodied agents”. This approach can cause serious issues concerning scalability, extensibility and performance of large-scale multi-robot systems. In this paper a different approach to design and implementation of software systems managing mobile robots is presented. It is based on the logical separation of a hardware robot and a software agent. The paper focuses on improvements in system performance achieved by using the approach. Three exemplary applications are described and experimental results are provided.

Extensible Web Crawler - Towards Multimedia Material Analysis

Methods of Web pages content monitoring come increasingly in the interest of law enforcement services, searching for Web pages contain symptoms of criminal activities. The information can be hidden from indexing systems by embedding in multimedia materials. Finding such materials is a large challenge of contemporary criminal analysis. A concept of integrating a large scale Web crawling system with a multimedia materials analysis algorithms is described in this paper. The Web crawling system, which is processing a few hundred pages per second, provides a mechanism for plugin inclusion. A plugin can analyze processed resources and detect references to multimedia materials. The references are passed to a component, which implements an algorithm for image or video analysis. Several approaches to the integration are described and some exemplary implementation assumptions are presented.

Generation-free Agent-based Evolutionary Computing

Abstract Metaheuristics resulting from the hybridization of multi-agent systems with evolutionary computing are efficient in many optimization problems. Evolutionary multi-agent systems (EMAS) are more similar to biological evolution than classical evolutionary algorithms. However, technological limitations prevented the use of fully asynchronous agents in previous EMAS implementations. In this paper we present a new algorithm for agent-based evolutionary computations. The individuals are represented as fully autonomous and asynchronous agents. Evolutionary operations are performed continuously and no artificial generations need to be distinguished. Our results show that such asynchronous evolutionary operators and the resulting absence of explicit generations lead to significantly better results. An efficient implementation of this algorithm was possible through the use of Erlang technology, which natively supports lightweight processes and asynchronous communication.

Robot Task Allocation Using Signal Propagation Model

Abstract The problem of Multi-Robot Task Allocation has received significant attention over last years. The solutions with decentralized decision making have proven better durability than those using centralized planning. In this paper a method which does not use any explicit communication is presented. It is especially suitable for very high dynamics of tasks priorities, when other methods must often perform time-consuming replanning. The method uses a simple model of signal emission and propagation in the environment. Simulation experiments are provided to demonstrate usability and interesting features of the method.

Extensible Multi-Robot System

In recent decades many important problems concerning mobile robots control algorithms have been solved. In order to integrate those solutions into complex, reliable, production-quality systems, an effort must be made to define abstract methodologies, architectures and frameworks, that would provide common functionalities and support features desirable in production systems: scalability, extensibility, flexibility and durability. This paper presents an agent-based architecture, which should provide extensibility of Multi-Robot Systems.

Erlang-based desynchronized urban traffic simulation for high-performance computing systems

Abstract The problem of simulating micro-scale urban traffic in large environments provides a great opportunity for the utilization of HPC systems. Parallel implementation of this kind of computation (where complex, data-intensive processing must be synchronized) is not straightforward. The simulation presented in this paper is based on the concept of controlled desynchronization of the computations, which does not violate the model. The implementation in the Erlang language uses Erlang distribution mechanisms for building and managing the computing cluster. The presented method makes it possible to simulate 11.5M cars traversing a road system composed of 240K crossroads and 144M possible locations. It provides a linear scalability of up to 19,200 computing cores (where simulation speeds reach 160 steps per second).

Highly scalable Erlang framework for agent-based metaheuristic computing

Abstract Difficult search and optimization problems, usually solved by metaheuristics, are very often implemented in concurrent and parallel environment, as many metaheuristics (e.g. population- or agent-based) are inherently easy to parallelize. Therefore search for easy-to-use, robust and efficient frameworks dedicated for such computing methods, especially in the era of ubiquitous many and multi-core systems, is very desirable. Indeed, the development of multi-core architectures is incredibly fast and multicore CPUs can be found nowadays not only in supercomputers, but also in ordinary laptops or even phones. Efficient use of multicore architectures requires applying suitable languages and technologies, like Erlang. Its concurrency model, based on lightweight processes and asynchronous message-passing, seems very well suited for running massively concurrent code on many cores. Most of existing Erlang industrial applications are deployed on computers with up to 24 CPU cores, and there are hardly any reports on using Erlang on architectures exceeding 32 physical cores. In this paper we present our experiences with developing a concurrent Erlang-based computing platform, scaling computationally-intensive and memory-intensive applications up to 64 cores, using as examples global optimization and urban traffic planning problems.

Towards Credible Driver Behavior Modeling

In order to predict the drivers behaviour, to adapt different mechanisms of the car or to tune up the traffic controls, one has to create a reliable drivers model. Such approaches have already been made, based on the assumption, that certain similarity must be observed for one driver doing the same maneuver (e.g. braking) or even in certain situations for different drivers. The research presented in this paper shows clearly that such assumptions are wrong in general, based on a number of real-life experiments using a real car (equipped with necessary sensors) in real traffic conditions. The experiments show that the same driver performs his/hers maneuvers differently, even though they could have been assumed to be the same or at least very similar. Thus necessity of further looking for reliable models of drivers behaviour is proven.

Extensible Platform for Studying the Behavior of Drivers in Urban Traffic

Research on traffic management and automation focuses on macro-scale problems involving large number of vehicles. It seems that the influence of individual decisions of particular drivers has not received sufficient attention over last decades. Understanding the characteristics of particular drivers is crucial for traffic optimization and for decisions making process performed by automated vehicles. In this paper we would like to present a prototype platform for studying characteristics of individual human drivers. The platform collects data during driving in urban environment and performs analysis and classification of observed behaviors. It is extensible in terms of measured parameters and classification methods. We also present results of first experiment using the created platform - classification of drivers during the left turn maneuver.