Search

In this paper we address the problem of automatically discovering optimal tactics in a combat scenario in which two opposing sides control a number of fighting units. Our approach is based on the evolution of behaviour trees, combined with simulation-based evaluation of solutions to drive the evolution. Our behaviour trees use a small set of possible actions that can be assigned to a combat unit, along with standard behaviour tree constructs and a novel approach for selecting which action from the tree is performed. A set of test scenarios was designed for which an optimal strategy is known from the literature. These scenarios were used to explore and evaluate our approach. The results indicate that it is possible, from the small set of possible unit actions, for a complex strategy to emerge through evolution. Combat units with different capabilities were observed exhibiting coordinated team work and exploiting aspects of the environment.

This article describes a set of proposed measures for characterizing areas within a virtual terrain in terms of their attributes and their relationships with other areas for incorporating game designers’ intent in gameplay requirement-based terrain generation. Examples of such gameplay elements include vantage point, strongholds, chokepoints and hidden areas. Our measures are constructed on characteristics of an isovist, that is, the volume of visible space at a local area and the connectivity of areas within the terrain. The calculation of these measures is detailed, in particular we introduce two new ways to accurately and efficiently calculate the 3D isovist volume. Unlike previous research that has mainly focused on aesthetic-based terrain generation, the proposed measures address a gap in gameplay requirement-based terrain generation – the need for a flexible mechanism to automatically parameterise specified areas and their associated relationships, capturing semantic knowledge relating to high level user intent associated with specific gameplay elements within the virtual terrain. We demonstrate applications of using the measures in an evolutionary process to automatically generate terrains that include specific gameplay elements as defined by a game designer. This is significant as this shows that the measures can characterize different gameplay elements and allow gameplay elements consistent with the designers’ intents to be generated and positioned in a virtual terrain without the need to specify low-level details at a model or logic level, hence leading to higher productivity and lower cost.

Behaviour trees offer a means to systematically decompose a behaviour into a set of steps within a tree structure. Genetic programming, which has at its core the evolution of tree-like structures, thus presents an ideal tool to identify novel behaviour patterns that emerge when the algorithm is guided by a set fitness function. In this paper, we present our framework for novel behaviour discovery using evolved behaviour trees, with some examples from the beyond-visual range air combat domain where distinct strategies emerge in response to modelling the effects of electronic warfare.

Software intensive Systems-of-Systems (SoS) are complex alliances of autonomous Constituent Systems (CSs) formed at a large scale to achieve a common objective. As such the CSs are operationally and managerially independent and geographically dispersed which generate emergent behaviors to achieve SoS missions through collective dynamics. Therefore, architectural modeling and analysis of a resulting SoS is pivotal to avoid stochastic architectural arrangements that can lead to undesired behaviors, systems outages, losses and non-conformance of core Quality Attributes (QAs) such as performance and reliability. In this research, we propose a formally founded approach for stochastic synthesis and automated verification of SoS architectural models to predict the impact of dynamic architectural changes on QAs at runtime. At first, we provide Hybrid Stochastic Formalism (HSF) based on Process Algebras (PAs) to model the stochastic SoS software architecture. At the architectural level, non-determinism is dealt with by treating HSF as Markov Decision Process (MDP). The SoS modeled with MDP is then verified against certain system properties using model checking through Probabilistic Computation Tree Logic (PCTL) operators. The effectiveness of our approach is evaluated through a fire monitoring and emergency response SoS to predict the impact of dynamic reconfiguration on QAs. The experimental results show that our method helps to assess different architectural configurations that support design choices to achieve missions without compromising quality.

Establishing reliable, clean, and inexpensive solar PV systems is a complex interplay between the level of reliability (LPSP), financial constraints, and CO2 emissions. This paper investigates the impact of these factors on stand-alone (SA) and grid-supplemented (GS) solar PV systems over multiple seasons. The research uses established hardware models, detailed power management strategies as well as realistic Australian grid tariffs and Genetic Algorithms to find the minimum Cost of Energy (COE) subject to LPSP and financial constraints. The developed power management strategies are also tested experimentally on a real solar PV system. The results indicate that the grid-supplemented system yields 30% lower COE compared to the stand-alone at baseline (LPSP

We present a software package for analysing body positions of a subject when they are lying or sleeping in their bed. The software is designed to interface to inexpensive sensors, such as the Microsoft Kinect, and is thus suitable for monitoring at the subjects own home, rather than a dedicated sleep lab. The system is invariant to bed clothing and levels of ambient lighting. Analysis time for a single night session is under five minutes, a significant improvement over the 30–60 min analysis time reported in the literature. Keywords: Lying position, Body position, Automated analysis

Concepts from the Australian mathematics curriculum on fractions were used as core elements to design three computer games. In each game, the concepts were presented in the form of tangible puzzles, customized to a difficulty level based on student capability. The games were integrated into a single virtual game world, and a fantasy story was used to help build a compelling experience. Five Year 6 classes were used to evaluate the game over four weeks. Three of the classes were provided with the games, and two served as a control. Both the intervention and control groups also covered fractions in class as part of the regular teaching program, consisting of instructor led content combined with access to online resources and activities. Participants completed a diagnostic test before the trial, and again at the end, designed to assess competence in the fractions concepts targeted by the game. Results show that on average students who had access to the game in addition to the regular teaching scored higher than control group students. In particular, looking at just students who started with a lower level of fractions skills, greater improvement was seen in those that had access to the game.

The automatic generation of behavioural models for intelligent agents in military simulation and experimentation remains a challenge. Genetic Algorithms are a global optimization approach which is suitable for addressing complex problems where locating the global optimum is a difficult task. Unlike traditional optimisation techniques such as hill-climbing or derivatives-based methods, Genetic Algorithms are robust for addressing highly multi-modal and discontinuous search landscapes. In this paper, we outline a simheuristic GA-based approach for automatic generation of finite state machine based behavioural models of intelligent agents, where the aim is the identification of novel combat tactics. Rather than evolving states, the proposed approach evolves a sequence of transitions. We also discuss workable starting points for the use of Genetic Algorithms for such scenarios, shedding some light on the associated design and implementation difficulties. Keywords: Simheuristics, Genetic algorithms, Multiagent simulations, Stochastic combinatorial optimization, Finite state machines

In this paper we introduce a novel approach to discovering emergent behaviour in multiagent simulations, using evolutionary finite state machines to model intelligent agents in an adversarial two-player game. Agent behaviour is modelled as a finite set of predetermined states. The logic that leads to transitions between states is evolved to maximise fitness, which is determined through execution in a constructive simulation environment. The resultant evolved finite state machine (E-FSM) is evaluated for two finite state machine implementations, one with states specifically designed to perform a known behaviour and the other with states consisting of generic actions. Our experiments demonstrate that this approach can discover complex emergent behaviours from simple, generic actions, and use these behaviours to achieve a position of tactical superiority in the domain of air combat simulation.

Action recognition using a combination of spatio-temporal skeleton based features.New data-driven methods for extracting discriminative features in each action class.A new similarity function based on LCSS for classification of new action instances.Evaluation of the approach using Kinect and mocap data showed improved performance.The approach achieved higher classification rate for noisy data in contrast to DTW. Human action analysis based on 3D imaging is an emerging topic. This paper presents an approach for the problem of action recognition using information from a number of action descriptors calculated from a skeleton fitted to the body of a tracked subject. In the proposed approach, a novel technique that automatically determines discriminative sequences of relative joint positions for each action class is employed. In addition, we use an extended formulation of the longest common subsequence algorithm as a similarity function, which allows the classifier to reliably find the best match for extracted features from noisy skeletal data. The proposed approach is evaluated using two existing datasets from the literature, one captured using a Microsoft Kinect camera and the other using a motion capture system. The experimental results show that the approach outperforms existing skeleton-based algorithms in terms of its classification accuracy and is more robust in the presence of noise when compared to the dynamic time warping algorithm for human action recognition.

Eliciting representative membership functions is one of the fundamental steps in applications of fuzzy theory. This paper investigates an unsupervised approach that incorporates variable bandwidth mean-shift and robust statistics for generating fuzzy membership functions. The approach automatically learns the number of representative functions from the underlying data distribution. Given a specific membership function, the approach then works out the associated parameters of the specific membership function. Our evaluation of the proposed approach consists of comparisons with two other techniques in terms of (i) parameterising MFs for attributes with different distributions, and (ii) classification performance of a fuzzy rule set that was developed using the parameterised output of these techniques. This evaluation involved its application using the trapezoidal and the triangular membership functions. Results demonstrate that the generated membership functions can better separate the underlying distributions and classifiers constructed using the proposed method of generating membership function outperformed three other classifiers that used different approaches for parameterisation of the attributes.

The recent interest in procedural content generation for video games has created the need to establish techniques for assessment of generated content. We present an investigation into the factors determining perceived difficulty in procedurally generated game levels. In doing so, an approach to identify relevant factors pertaining to player experience is established, which is subsequently used in the development of predictive difficulty models. In this paper, we apply our methodology to the genre of 2D platformers, presenting an investigation into factors related to difficulty, the development of a test-bed that can be used to collect the data, data collection and subsequent analysis. We investigate the contribution of the identified game and player metrics towards predicting difficulty using Multi-Layer Perceptron, J48 and Random Forest classifiers from WEKA. This work is presented as a preliminary investigation into modeling difficulty from procedural content. Significantly, this investigation provides a preliminary insight into metrics that can be used for developing a classification model for perceived difficulty.

The virtual terrain for a video game generally needs to exhibit a collection of gameplay elements, such as some areas suitable for hiding and others for large scale battles. A key problem in automating terrain design is the lack of a quantitative definition of terrain gameplay elements. In this paper, we address the problem by proposing a representation for gameplay elements based on a combination of space-based isovist measures from the field of architecture and graph-connectivity metrics. We then propose a genetic algorithm-based approach that evolves a set of modifications to an existing terrain so as to exhibit the gameplay element characteristics. The potential for this approach in the design of computer game environments is examined by generating terrain containing instances of the “hidden area” game element type. Results from four preliminary tests are described to show the potential of this research.

This article describes “WAMbot,'' an autonomous unmanned ground vehicle and the Team MAGICian solution to the 2010 Multi Autonomous Ground-robotic International Challenge (MAGIC). The system architecture, hardware, sensing, planning, navigation, mapping systems, and user interface were specifically designed to address the demands of the large-scale distributed reconnaissance and surveillance task presented in the challenge. The resulting system was selected as a finalist and ultimately placed 4th in MAGIC 2010. © 2012 Wiley Periodicals, Inc. © 2012 Wiley Periodicals, Inc.

This paper presents a novel approach for automatic detection of abnormal behaviours in daily routine of people living alone in their homes, without any manual labelling of the training dataset. Regularity and frequency of activities are monitored by estimating the status of specific electrical appliances via their power signatures identified from the composite power signal of the house. A novel unsupervised clustering technique is presented to automatically profile the power signatures of electrical devices. Then, the use of a test statistic is proposed to distinguish power signatures resulted from the occupant interactions from those of self-regulated appliances such as refrigerator. Experiments on real-world data showed the effectiveness of the proposed approach in terms of detection of the occupant's interactions with appliances as well as identifying those days that the behaviour of the occupant was outside the normal pattern.

In this paper, an approach is presented for the detection of abnormal behaviours in the activities of people living alone in their homes. The proposed approach takes input from Kinect cameras and extracts a number of visual attributes that represent the occupant's location and orientation. A set of fuzzy logic parameters is first learnt from the training data. Next the proposed approach learns epochs of activities in each location and then generates models of normal behaviour patterns. Unusual behaviour is then detected in subsequent data by looking for patterns which differ from the learnt normal behaviours based on their time of occurrence, visual attributes, or duration. Experiments conducted showed the effectiveness of the proposed system.

The rapid growth of the entertainment industry has driven the requirement for more efficient development of computerized games. Importantly, the diversity of audiences that participate in playing games has called for the development of new technologies that allow games to address users with differing levels of skills and preferences. This study explored the application of dynamic difficulty adjustment in computerized 2D plat former games through procedural level generation and interactive evolutionary computation. Computerized agents were introduced into the evolutionary cycle to play game levels in the player's stead. Additionally, agent behavior was based on the player's game-play characteristics, which has provided a more practical means for evaluating and evolving game levels tailored to a particular player's skills.

This paper introduces a new approach to the procedural generation of maze-like game level layouts by evolving CA. The approach uses a GA to evolve CA rules which, when applied to a maze configuration, produce level layouts with desired maze-like properties. The advantages of this technique is that once a CA rule set has been evolved, it can quickly generate varying instances of maze-like level layouts with similar properties in real time.

A problem in configuring modular self-reconfigurable robots with distributed control is determining whether a module can be moved without breaking the system apart. In finding a set of modules that can be safely moved, the key issues are to maximize the number of mobile modules identified and minimize the time consumed in finding these modules. In this study, a distributed search algorithm is proposed and compared to an existing approach.

The increasing elderly population presents a challenge on the resources of carers and assisted living communities. In this paper, we present an algorithm based around the Microsoft Kinect for monitoring activities of daily living. The system analyses the behaviour of occupants to provide carers with valuable observational data, and has the capacity to detect abnormal events in the home.

The addition of an exercise component to computer games is a method that attempts to increase exercise compliance among the population. Various attempts at such games have been made, both commercially and as research projects, but the set of studied game mechanics that are both entertaining and provide beneficial exercise are relatively small. In our work, a game mechanic was designed for a floor-based input controller (DDR Pad). Rather than the standard mechanic usually employed for this device, our approach allows the player to explore a virtual world, requiring exertive movements on the input device in order to progress. The exercise is motivated implicitly through the game goals and a mechanism exists to control the pace of the game by monitoring the player's heart rate. In our trials, we have found that the game is engaging for the participants. The game also provides a beneficial level of exercise, with the player's heart rate, on average, rising to a level that constitutes moderate exercise after three minutes and forty seconds, and increasing from thereon.

Coevolutionary algorithms are a special kind of evolutionary algorithm with advantages in solving certain specific kinds of problems. In particular, competitive coevolutionary algorithms can be used to study problems in which two sides compete against each other and must choose a suitable strategy. Often these problems are multimodal --- there is more than one strong strategy for each side. In this paper, we introduce a scalable multimodal test problem for competitive coevolution, and use it to investigate the effectiveness of some common coevolutionary algorithm enhancement techniques in improving a coevolutionary algorithm's ability to explore landscapes with multiple optima.

In this study, we introduce MC-TSAR, a Monte Carlo Tree Search algorithm for strategy selection in simultaneous multistage games. We evaluate the algorithm using a battle planning scenario in which replanning is possible. We show that the algorithm can be used to select a strategy that approximates a Nash equilibrium strategy, taking into account the possibility of switching strategies part way through the execution of the scenario in the light of new information on the progress of the battle.

Coevolutionary algorithms are a special kind of evolutionary algorithm with advantages in solving certain specific kinds of problems. In particular, competitive coevolutionary algorithms can be used to study problems in which two sides compete against each other and must choose a suitable strategy. Often these problems are multimodal -- there is more than one strong strategy for each side. In this paper, we introduce a scalable multimodal test problem for competitive coevolution, and use it to investigate the effectiveness of some common coevolutionary algorithm enhancement techniques.

In this paper, we investigate the effectiveness of several techniques commonly recommended for overcoming convergence problems with coevolutionary algorithms. In particular, we investigate effects of the Hall of Fame, and of several diversity maintenance methods, on a problem designed to test the ability of coevolutionary algorithms to deal with an intransitive superiority relation between solutions. We measure and analyse the effects of these methods on population diversity and on solution quality.

Exergaming, the merger of exercise and video games, tries to use the engaging experience of playing a video game to help people achieve their exercise requirements. To guide the design of such games the dual flow model, an extension of the theory of flow to both mental and physical experience, has been proposed. This paper presents the development of an exergame system designed to demonstrate the validity of the dual flow model, along with initial results from a pilot trial. The results show that such a game system can be used to deliver the required exercise across a range of participants.

Doppler ultrasound monitors are commonly used to measure the Fetal Heart Rate (FHR) in routine screenings. They are less intrusive than Fetal Electrocardiography (FECG), but are more prone to noise and thus seen as less accurate devices. Recent developments resulted in beat-to-beat heart rate algorithms for Doppler shift monitors that claimed to approach the accuracy of that obtained from direct FECG. We evaluate one such algorithm, along with a noise tolerant improvement using a mixture of synthetic and real fetal heart data in order to quantify their accuracy and robustness in the presence of the typical noise in signals from an ultrasound probe. The findings showed that our proposed approach achieved higher accuracy and better robustness compared to the beat-to-beat fetal heart detection approach.

It has been shown that the rate of preterm birth in remote and rural populations in Australia is more than twice of those in the metropolitan population, and that increased access to quality antenatal care in such communities leads to a significant improvement. This paper describes a system that supports remote fetal heart rate monitoring, thus providing a technique that addresses this need for better remote antenatal care. The application involves a smart phone, a portable Doppler ultrasound device, a remote back-end server and network connectivity provided by the Global System for Mobile communications (GSM) and General Packet Radio Service (GPRS) networks. Algorithms for fetal heart rate detection were adapted to cater for the lower processing powers on the smart phone. The accuracy of this detection algorithm in the mobile application was evaluated and found to be highly comparable to that of a MATLAB implementation on a PC.

The scanning and computerized processing of images hadits birth in 1956 at the National Bureau of Standards (NBS,nowNationalInstituteofStandardsandTechnology(NIST))[1].Image enhancementalgorithms weresome of the first tobe developed [2]. Half a century later, literally thousands ofimage processing algorithms have been published. Some ofthese have been specific to certain applications such as theenhancement of latent fingerprints, whilst others have beenmore generic in nature, applicable to all, yet master of none.The scope of these algorithms is fairly expansive, rangingfrom automatically extracting and delineating regions of in-terest such as in the case of segmentation, to improving theperceived quality of an image, by means of image enhance-ment. Since the early years of image processing, as in manysubfields of software design, there has been a portion of thedesign process dedicated to algorithm testing. Testing is theprocessofdeterminingwhetherornotaparticularalgorithmhassatisfieditsspecificationsrelatingtocriteriasuchasaccu-racy and robustness. A major limitation in the design of im-ageprocessingalgorithmsliesinthedifficultyindemonstrat-ingthatalgorithmsworktoanacceptablemeasureofperfor-mance. The purpose of algorithm testing is two-fold. Firstlyit provides either a qualitative or a quantitative method ofevaluating an algorithm. Secondly, it provides a comparativemeasure of the algorithm against similar algorithms, assum-ing similar criteria are used. One of the greatest caveats indesigning algorithms incorporating image processing is howtoconceivethecriteriausedtoanalyzetheresults.Dowede-signacriterionwhichmeasuressensitivity,robustness,orac-curacy? Performance evaluation in the broadest sense refersto a measure of some required behavior of an algorithm,whether it is achievable accuracy, robustness, or adaptabil-ity. It allows the intrinsic characteristics of an algorithm tobe emphasized, as well as the evaluation of its benefits andlimitations.More often than not though, such testing has been lim-ited in its scope. Part of this is attributable to the actual lackof formal process used in performance evaluation of im-age processing algorithms, from the establishment of testingregimes, to the design of metrics. Selection of an appropri-ate evaluation methodology is dependent on the objectiveof the task. For example, in the context of image enhance-ment, requirements are essentially different for screen-basedenhancementandenhancementwhichisembeddedwithinasubalgorithm. Screen-based enhancement is usually assessedinasubjectivemanner,whereaswhenanalgorithmisencap-sulated within a larger system, subjective evaluation is notavailable,andthealgorithmitselfmustdeterminethequalityof a processed image. Very few approaches to the evaluationofimageprocessingalgorithmscanbefoundintheliterature,although the concept has been around for decades. A signif-icant difficulty which arises in the evaluation of algorithmsis finding suitable metrics which provide an objective mea-sure of performance. A performance metric is a meaningfuland computable measure used for quantitatively evaluatingthe performance of any algorithm. Consider the process ofassessing image quality. There is no single quantitative met-ric which correlates well with image quality as perceived bythe human visual system. The process of analyzing failure isintrinsically coupled with the process of performance evalu-ation.Inordertoascertainwhetheranalgorithmfailsornot,youhavetodefinethecharacteristicsofsuccess.Failureanal-ysisistheprocessofdeterminingwhyanalgorithmfailsdur-ing testing. The knowledge generated is then fed back to thedesign process in order to engender refinements in the algo-rithm.Thisisadifficultprocessinapplicationssuchasimageenhancement primarily because there is usually no referenceimagewhichcanbeusedasan“ideal”image.Theassessmentofimagequalityplaysanimportantroleinapplicationssuchas consumer electronics. Metrics could be used to monitororoptimizeimagequalityindigitalcameras,benchmarkandevaluate image enhancement algorithms. There is no singlemetric that correlates well with image quality as perceivedby the human visual system. Selection of an appropriate

In this paper, we report on our experiences in using an animated competitive game with simulated physics to teach Artificial Intelligence techniques in an undergraduate computer science course. Students develop intelligent controllers for simulated vehicles, which compete with each other in a tournament. The simulation includes a real-time visualisation of the contests, and the students' solutions utilise an AI toolkit that provides animated displays showing the internal workings of their controllers in parallel with the simulation. The result is a learning experience that engages students' enthusiasm, and helps them to development mental models of the AI algorithms.

A threshold-based algorithm is presented for the extraction of the pectoral muscle edge in mediolateral oblique view mammograms. The minimum cross-entropy thresholding algorithm is applied to local areas around the pectoral muscle to determine a series of thresholds as a function of area size. Using a model image it is shown that art inflection point in this function corresponds to a threshold that will separate the pectoral muscle from the rest of the breast. Post processing is performed on mammograms to eliminate false positive points of inflection and a straight line is fitted to the detected pectoral boundary in order to smooth jaggedness caused by the non-uniform intensity of the pectoral muscle edge.

A framework for simulating physiological responses to exercise lets exergame developers more efficiently test their games during development. An example implementation combines software that simulates heart-rate responses with a custom hardware setup. Simulated heart-rate responses closely match the real physiological responses, demonstrating the approach's validity and potential.