Search

Proposes to study the absolute periodicity of delayed neural networks. A neural network is said to be absolutely periodic, if for every activation function in some suitable functional set and every input periodic vector function, a unique periodic solution of the network exists and all other solutions of the network converge exponentially to it. Absolute stability of delayed neural networks is also studied in this paper. Simple and checkable conditions for guaranteeing absolute periodicity and absolute stability are derived. Simulations for absolute periodicity are given.

The chapter explains the computational intelligence techniques utilized in the algorithms presented in the book. The fuzzy and rough sets, fuzzy-rough sets, genetic algorithm and, feature selection and classification using the fuzzy-rough sets are detailed. The biologically inspired feature extraction system utilized in the presented algorithms is explained.

The boosting-based fuzzy-rough multi-cluster (BFRMC) classification algorithm utilizing the rough set concepts of fuzzy lower and upper approximations is discussed in this chapter. The BFRMC algorithm transforms each quantitative value of a feature into fuzzy sets of linguistic terms using membership functions and calculates the fuzzy lower and upper approximations. The membership functions are generated from cluster points generated by the subtractive clustering technique. A certain rule set based on fuzzy lower approximation and a possible rule set based on fuzzy upper approximation are generated. Iterative rule learning based genetic algorithm is employed in combination with a boosting technique, for generating the possible rule set, incrementally optimizing each fuzzy classifier rule. The BFRMC classifier’s performance on the datasets from the UCI machine learning repository is discussed.

A fuzzy-rough multi cluster (FRMC) classifier for the recognition of hand postures and face is presented in this chapter. Features of the image are extracted using the computational model of the ventral stream of visual cortex. The recognition algorithm translates each quantitative value of the feature into fuzzy sets of linguistic terms using membership functions. The membership functions are generated by the fuzzy partitioning of the feature space into fuzzy equivalence classes, using the feature cluster centers generated by the subtractive clustering technique. A rule base generated from the lower and upper approximations of the fuzzy equivalence classes classifies the images through a voting process. Using Genetic Algorithm (GA), the number of features required for classification is reduced by identifying the predictive image features. The margin of classification, which is a measure of the discriminative power of the classifier, is used to ensure the quality of classification process. The fitness function suggested assists in the feature selection process without compromising on the classification accuracy and margin. The algorithm is tested using two hand posture and three face datasets. The algorithm provides good classification accuracy, at a less computational effort. The selection of relevant features further reduced the computational costs of both feature extraction and classification algorithms, which makes it suitable for real-time applications. The performance of the algorithm is compared with that of Support Vector Machines.

An overview of the visual pattern recognition process and associated key issues are presented in this chapter. The varying scales and shapes, inter-class similarity, large number of features, and complex backgrounds are issues related to visual pattern recognition. The book focuses on these issues. The chapter introduces different algorithms addressing these issues.

This chapter focuses on feature selection and classification of multi-feature patterns. Micro array based cancer classification and image based face recognition are discussed. A detailed review of hand gesture recognition algorithms and techniques is included. The hand gesture recognition algorithms are surveyed by classifying them into three categories (a) hidden Markov model based methods, (b) neural network and learning based methods, and (c) the other methods. A list of available hand gesture databases is provided.

Classification of datasets with multiple features is computationally intensive. Fuzzy-rough set based feature selection and classification requires reduced computational efforts. Lower and upper approximations of fuzzy equivalence classes are useful in finding discriminative features and classification boundaries in a dataset. This chapter discusses a fuzzy-rough single cluster (FRSC) classifier which is a discriminative feature selection and classification algorithm. The FRSC classifier translates each quantitative value of a feature into fuzzy sets of linguistic terms using membership functions and, identifies discriminative features. The membership functions are formed by partitioning the feature space into fuzzy equivalence classes, using feature cluster centers identified through subtractive clustering. Classification rules are generated using fuzzy membership values partitioning the lower and upper approximations. The patterns are classified through a voting process. Both the feature selection and classification algorithms have polynomial time complexity. The algorithm is tested in two types of classification problems, namely, cancer and image-pattern classification. The large number of gene expression profiles and relatively small number of available samples make the feature selection a key step in microarray based cancer classification. The algorithm identified relevant features (predictive genes in the case of cancer data) and provided good classification accuracy, at a less computational cost, with good margin of classification. A comparison of the performance of the FRSC classifier with other relevant classification methods shows the classifier’s better discriminative power.

The Attention based Segmentation and Recognition (ASR) algorithm for hand postures against complex backgrounds is discussed in this chapter. The ASR algorithm can detect, segment and recognize multi-class hand postures. Visual attention, which is a cognitive process of selectively concentrating on a region of interest in visual field, helps humans to recognize objects in cluttered natural scenes. The ASR algorithm utilizes a Bayesian model of visual attention to generate a saliency map, and to detect and identify the hand region. Feature based visual attention is implemented using a combination of high level (shape, texture) and low level (color) image features. The shape and texture features are extracted from a skin similarity map, using a computational model of the ventral stream of visual cortex. The skin similarity map, which represents the similarity of each pixel to the human skin color in HSI color space, enhances the edges and shapes within the skin colored regions. The color features used are discretized chrominance components in HSI, YCbCr color spaces, and similarity-to-skin map. The hand postures are classified using shape and texture features, with a support vector machines classifier. The NUS hand posture dataset-II with 10 classes of complex background hand postures is utilized for testing the algorithm. The dataset contains hand postures from 40 subjects of different ethnicities. A total of 2,750 hand postures and 2,000 background images are available in the dataset. The hand postures vary in size and shape. The ASR algorithm is tested for hand detection and hand posture recognition using 10 fold cross-validation. The experimental results show that the algorithm has a person independent performance, and is reliable against variations in hand sizes and complex backgrounds.

© 2014 Elsevier B.V.In robotics, lower-level controllers are typically used to make the robot solve a specific task in a fixed context. For example, the lower-level controller can encode a hitting movement while the context defines the target coordinates to hit. However, in many learning problems the context may change between task executions. To adapt the policy to a new context, we utilize a hierarchical approach by learning an upper-level policy that generalizes the lower-level controllers to new contexts. A common approach to learn such upper-level policies is to use policy search. However, the majority of current contextual policy search approaches are model-free and require a high number of interactions with the robot and its environment. Model-based approaches are known to significantly reduce the amount of robot experiments, however, current model-based techniques cannot be applied straightforwardly to the problem of learning contextual upper-level policies. They rely on specific parametrizations of the policy and the reward function, which are often unrealistic in the contextual policy search formulation. In this paper, we propose a novel model-based contextual policy search algorithm that is able to generalize lower-level controllers, and is data-efficient. Our approach is based on learned probabilistic forward models and information theoretic policy search. Unlike current algorithms, our method does not require any assumption on the parametrization of the policy or the reward function. We show on complex simulated robotic tasks and in a real robot experiment that the proposed learning framework speeds up the learning process by up to two orders of magnitude in comparison to existing methods, while learning high quality policies.

This paper explores how stimulus-dependent habituation can be integrated to modulate a cognitively inspired bottom up learning model of semantic representation. We first describe the equations governing habituation and associative habituation and describe how to integrate it to a bottom up hierarchical habituation model. Habituation is used to modulate the weights between the parent and children in the hierarchy. We investigate the behavior of habituation in the hierarchy, the properties of habituation and recovery in the model and show that while habituation has no detrimental effect on clustering, it reduces the number of neurons required for representation and therefore propose it as an important component of cognitively inspired memory systems.

An elastic graph matching algorithm using biologically inspired features is proposed for the recognition of hand postures. Each node in the graph is labeled using an image feature extracted using the computational model of the ventral stream of visual cortex. The graph nodes are assigned to geometrically significant positions in the hand image, and, the model graphs are created. Bunch graph method is used for modeling the variability in hand posture appearance. Recognition of a hand posture is done by the elastic graph matching between the model graphs and the input image. A radial basis function is used as the similarity function for the matching process. The proposed algorithm is tested on a 10 class hand posture database which consists of 478 grey scale images with light and dark backgrounds. The algorithm provided better recognition accuracy (96.35%) compared to the reported results (93.77%) in the literature.

Over the past decades, the design and development of mission based Autonomous Underwater Vehicle (AUV) continues to challenge researchers. Although AUV technology has matured and commercial systems have appeared in the market, a generic yet robust AUV command and control (C2) system still remains a key research area. This paper presents a command and control system architecture for modular AUVs. We particularly focus on the design and development of a generic control and software architecture for a single modular AUV while allowing natural extensions to multi-vehicle scenarios. This proposed C2 system has a hybrid modular-hierarchical control architecture. It adopts top-down approach in mission level decision making and task planning while utilizing bottom-up approach for navigational control, obstacle avoidance and vehicle fault detection. Each level consists of one or more autonomous agent components handling different C2 tasks. This structure provides the vehicle developers with an explicit view of the clearly defined control responsibilities at different level of control hierarchy. The resultant C2 system is currently operational on the STARFISH AUV built at the ARL of the National University of Singapore. It has successfully executed some autonomous missions during sea trials carried out around the Singapore coastal area.

In this paper, we present an approach to motor skill learning based on internal models. By pursuing the temporal and spatial scalability of internal models, we first investigate the possibility of generating similar movement patterns directly via the same internal model with the minimum changes in the internal model parameters, and avoid the reinforcement learning. Next, we consider more complex movements for which different internal models are needed. Based on the task decomposition, all movements can be classified into the sequential and parallel DMPs. The former requires a number of IMs to work sequentially so that a sophisticated motor behavior can be performed. The latter also requires a number of IMs to work in parallel to generate the needed movement patterns. To mimic the human limb behavior, a two-link robot arm is used as the first prototype to perform the motor learning process of letter writing. A FUJITSU HOAP-1 humanoid robot is used as the second prototype and the upper limb movement is conducted in real-time, which further validates the effectiveness of multiple internal model approach for motor learning.

The Gyrobot is a gyroscopically stabilized, single-wheeled robot. A fast spinning flywheel mounted inside the frame provides lateral stability of the wheel-shaped robot. Steering of the robot is effectuated by tilting the fly-wheel with the help of a tilt motor, and forward/backward motion is implemented by another motor mounted on the heavy platform. This structure of robot shows excellent dynamic stability and is, therefore, suitable for navigation in unknown, rough terrain. From this point of view, the Gyrobot has definite advantage over other types of wheeled robots. Moreover, the electronic components are placed inside the sealed shell of the wheel giving it an advantage in operations where navigation in wet surfaces is required, for example inspection of sewerage lines or chemical spills. In this paper, we report the design and fabrication of the Gyrobot and present the results of path following navigating algorithm designed for this robot. A linear feedback controller is adopted at this stage to control the tilt and forward/backward motion of the robot.

The parallelism in DNA computing provides great potential to solve complex problems. DNA coding methods are useful to promote the genetic algorithms' (GAs) performance on some problems. This work compares two coding methods for the genetic algorithm. The effect of the developed coding methods on the algorithms is analyzed by application to classification problem.

The gyrobot is a gyroscopically stabilized, single-wheeled robot. A fast spinning flywheel mounted inside the frame provides lateral stability of the wheel-shaped robot. Steering of the robot is effectuated by tilting the flywheel with the help of a tilt motor, and forward/backward motion is implemented by another motor mounted on the heavy platform. This structure of robot shows excellent dynamic stability and is, therefore, suitable for navigation in unknown, rough terrain. From this point of view, the gyrobot has definite advantage over other types of wheeled robots. Moreover, the electronic components are placed inside the sealed shell of the wheel giving it an advantage in operations where navigation in wet surfaces is required, for example inspection of sewerage lines or chemical spills. In this paper, we report the design and fabrication of the gyrobot and present the results of path following navigating algorithm designed for this robot. A linear feedback controller is adopted at this stage to control the tilt and forward/backward motion of the robot.

In recent years, the new concept of DNA (deoxyribonucleic acid) computing has drawn intensive research interests. The idea of DNA computing, proposed by Leonard Adleman (Leonard 1994) in 1994, is to express a problem in the form of DNA molecules and to realize the computation by operating on those DNA molecules. There are two major advantages of DNA computing: the great parallel computation power and the mega information storage ability. DNA computing is quick, as it can perform many calculations simultaneously or in parallel (Boneh et al 1995; Winfree 1995). Some of the very complex problems which are hard even for supercomputers can be solved by DNA computing (Lipton 1995; Boneh et al 1995). DNA computing also provides a huge storage media since it stores the information in DNA molecules (Baum 1995). DNA computing is such a novel idea that its future applications still remain unknown. However, it seems that DNA computing will make great changes in the fields of computer science, biology, chemistry and medicine (Leonard 1996; Beaver 1995).

The advantages and disadvantages of evolving neural control systems for mobile robots using genetic algorithms are investigated. The Khepera robot is trained using the evolutionary neural networks (ENN) algorithm for the task of obstacle avoidance. The feasibility of using Q-learning for robot learning is also studied. It is found that Q-learning can be successfully used to train a robot and is more promising than the ENN algorithm in this case. The Webots simulation software has been used to carry out all the experiments.

A comparison of mobile robot navigation using evolutionary neural networks and the pain based algorithm is discussed in this paper. The controllers are designed based on evolutionary neural networks and the pain-based algorithms. The performance of the controllers are verified with the Khepera robot.

This chapter discusses the application of the evolutionary artificial potential field (EAPF) in mobile robot path planning. The parameters of the evolutionary artificial potential field are optimized with the multi-objective evolutionary algorithm. The EAPF is utilized in a robot soccer system.

Evolutionary artificial potential field (EAPF) functions are utilized for mobile robot navigation in a microrobot soccer (MiroSot) environment. In a micro-robot soccer system the robots are monitored using an overhead CCD Camera, making it suitable for real time application of the EAPF functions. The effectiveness of the EAPF functions in real time mobile robot navigation are verified through experimentation. The EAPF functions proposed were tested in different scenarios related to ball tracking and ball kicking, while facing competition from other robots.

A DNA coded genetic-algorithm (GA) is proposed to optimize the rule-base of a fuzzy logic controller (FLC). The controller is designed for a vehicle-active suspension system to improve the driving comfort. The DNA codes GA constructed optimal decision-making rules for the fuzzy logic controller. Simulation results demonstrate the effectiveness of the algorithm.

A new methodology named Evolutionary Artificial Potential Field (EAPF) is proposed for real-time robot path planning. The artificial potential field method is combined with genetic algorithms, to derive optimal potential field functions. The proposed EAPF approach is capable of navigating robot(s) situated among moving obstacles. Potential field functions for obstacles and goal points are also defined. The potential field functions for obstacles contain tunable parameters. The multi-objective evolutionary algorithm (MOEA) is utilized to identify the optimal potential field functions. Fitness functions such as goal-factor, obstacle-factor, smoothness-factor and minimum-pathlength-factor are developed for the MOEA selection criteria. An algorithm named escape-force is introduced to avoid the local minima associated with EAPF. Moving obstacles and moving goal positions were considered to test the robust performance of the proposed methodology. Simulation results show that the proposed methodology is efficient and robust for robot path planning with non-stationary goals and obstacles.

In this study, a novel method, process noise identification-based particle filter is proposed for tracking highly manoeuvring target. In the proposed method, the equivalent-noise approach is adopted, which converts the problem of manoeuvring target tracking to that of state estimation in the presence of non-stationary process noise with unknown statistics. A novel method for identifying the non-stationary process noise is proposed in the particle filter framework. Compared with the multiple model approaches for manoeuvring target tracking, the proposed method needs to know neither the possible multiple models nor the transition probability matrices. One simple dynamic model is adopted during the whole tracking process. The proposed method is especially suitable for tracking highly manoeuvring target because of its capability of dealing with sample impoverishment, which is a common problem in particle filter and becomes serious when tracking large uncertain dynamics.

Multi-agent architectures for autonomous robots are generally mission and platform oriented. Autonomous robots are commonly employed in patrolling, surveillance, search and rescue and human-hazardous missions. Irrespective of the differences in unmanned aerial and ground robots, the algorithms for obstacle detection and avoidance, path planning and path-tracking can be generalized. Service-oriented interoperable framework for robot autonomy (SOIFRA) proposed in this paper is an interoperable multi-agent framework focusing on generalizing platform-independent algorithms for unmanned aerial and ground vehicles. As obstacle detection and avoidance are standard requirements for autonomous robot operation, platform-independent collision avoidance algorithms are incorporated into SOIFRA. SOIFRA is behaviour based and is interoperable across unmanned aerial and ground vehicles. Obstacle detection and avoidance are performed utilizing computer vision-based algorithms, as these are generally platform independent. Obstacle detection is achieved utilizing Hough transform, Canny contour and Lucas–Kanade sparse optical flow algorithm. Collision avoidance performed utilizing optical flow-based and expansion of object-based time-to-contact demonstrates SOIFRA’s modularity. Experiments performed, utilizing TurtleBot, Clearpath Robotics Husky, AR Drone and Hector-quadrotor, establish SOIFRA’s interoperability across several robotic platforms.

In this paper, a novel method, process noise identification based particle filter is proposed for tracking highly maneuvering target. In the proposed method, the equivalent-noise approach [1], [2], [3] is adopted, which converts the problem of maneuvering target tracking to that of state estimation in the presence of non-stationary process noise with unknown statistics. A novel method for identifying the non-stationary process noise is proposed in the particle filter framework. Compared with the multiple model approaches for maneuvering target tracking, the proposed method needs to know neither the possible multiple models nor the transition probability matrices. One simple dynamic model is adopted during the whole tracking process. The proposed method is especially suitable for tracking highly maneuvering target due to its capability of dealing with sample impoverishment, which is a common problem in particle filter and becomes serious when tracking large uncertain dynamics.

In this paper, the particle filter is used to solve the nonlinear and nonGaussian estimation problem in multiple targets tracking and multiple sensor fusion process. The weight of the particle is evaluated through the combination of Joint Probability Data Association (JPDA) and multiple hypothesis tracking (MHT), which makes the probabilistic assignment based on all reasonable hypotheses in a sliding window of multiple scans. To track the multiple targets with random varying velocities, each particle's state is optimized based on the history information from the previous scans in the sliding window and group information in the current scan. The particle diversity is enriched while the trajectory of each particle evolves towards the high posterior density distribution. Moreover the problem of tracking newly appeared objects or disappeared objects are also discussed. The simulation results show that the improved particle filter method achieves dynamic stability and robustness while tracking multiple random moving targets.

International audience; The essence of humanoid robots is their ability to reproduce human skills in locomotion and manipulation. Early efforts in humanoid research were dedicated to bipedal walking, first on flat terrains and recently on uneven ones, while the manipulation capabilities inherit from the literature in bimanual and dexterous-hand manipulation. In practice, the two problems interact largely. Locomotion in cluttered spaces benefits from extra contacts between any part of the robot and the environment, such as when grippers grasp a handrail during stair climbing, while legs can conversely enhance manipulation capabilities, such as when arching the whole body to augment contact pressure at an end effector. The two problems share the same background: they are governed by non-smooth dynamics (friction and impacts at contacts) under viability constraints including dynamic stability. Consequently, they are now solved jointly. This chapter highlights the state-of-the-art techniques used for this purpose in multi-contact planning and control.