International Journal of Humanoid Robotics

A novel algorithm based on fuzzy-rough sets is proposed for the recognition of hand postures and face. 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 formed 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, 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 proposed algorithm is tested using two hand posture and three face datasets. The algorithm provided 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 proposed algorithm is compared with that of support vector machines.

Robust vision in dynamic environments using limited processing power is one of the main challenges in robot vision. This is especially true in the case of biped humanoids that use low-end computers. Techniques such as active vision, context-based vision, and multi-resolution are currently in use to deal with these highly demanding requirements. Thus, having as main motivation the development of robust and high performing robot vision systems, which can operate in dynamic environments, with limited computational resources, we propose a spatiotemporal context integration framework that improves the perceptual capabilities of a given robot vision system. Furthermore, we try to link the vision, tracking, and self-localization problems using a context filter to improve the performance of all these parts together more than to improve them separately. This framework computes: (i) an estimation of the poses of visible and nonvisible objects using Kalman filters; (ii) the spatial coherence of each current detection with all other simultaneous detections and with all tracked objects; and (iii) the spatial coherence of each tracked object with all current detections. Using a Bayesian approach, we calculate the a-posteriori probabilities for each detected and tracked object, which is used in a filtering stage. We choose as a first application of this framework, the detection of static objects in the RoboCup Standard Platform League domain, where Nao humanoid robots are employed. The proposed system is validated in simulations and using real video sequences. In noisy environments, the system is able to decrease largely the number of false detections and to improve effectively the self-localization of the robot.

We present a novel representation of visual information, based on local symbolic descriptors, that we call visual primitives. These primitives: (1) combine different visual modalities, (2) associate semantic to local scene information, and (3) reduce the bandwidth while increasing the predictability of the information exchanged across the system. This representation leads to the concept of early cognitive vision that we define as an intermediate level between dense, signal-based early vision and high-level cognitive vision. The framework's potential is demonstrated in several applications, in particular in the area of robotics and humanoid robotics, which are briefly outlined.

An original method to build a visual model for unknown objects by a humanoid robot is proposed. The algorithm ensures successful autonomous realization of this goal by addressing the problem as an active coupling between computer vision and whole-body posture generation. The visual model is built through the repeated execution of two processes. The first one considers the current knowledge about the visual aspects and the shape of the object to deduce a preferred viewpoint with the aim of reducing the uncertainty of the shape and appearance of the object. This is done while considering the constraints related to the embodiment of the vision sensors in the humanoid head. The second process generates a whole robot posture using the desired head pose while solving additional constraints such as collision avoidance and joint limitations. The main contribution of our approach relies on the use of different optimization algorithms to find an optimal viewpoint by including the humanoid specificities in terms of constraints, an embedded vision sensor, and redundant motion capabilities. This approach differs significantly from those of traditional works addressing the problem of autonomously building an object model.

In this article, we propose a conformal model for 3D visual perception. In our model, the two views are fused in an extended 3D horopter model. For visual simultaneous localization and mapping (SLAM), an extended Kalman filter (EKF) technique is used for 3D reconstruction and determination of the robot head pose. In addition, the Viola and Jones machine-learning technique is applied to improve the robot relocalization. The 3D horopter, the EKF-based SLAM, and the Viola and Jones machine-learning technique are key elements for building a strong real-time perception system for robot humanoids. A variety of interesting experiments show the efficiency of our system for humanoid robot vision.

In this work, an explicitly task-oriented approach to the active vision problem is presented. The system tries to reduce the most relevant components of the uncertainty in the world model, for the task the robot is currently performing. It is task oriented in the sense that it explicitly considers a task-specific value function. As test-bed for the presented active vision approach, we selected a robot soccer attention problem: goal-covering by a goalie player. The proposed system is compared with information-based approaches. Experimental results show that it surpasses them in the tested application. We conclude that, when the goal is not the uncertainty reduction itself, the minimization of the belief entropy is not a useful optimality criterion, and that for such cases, task-oriented optimality criteria are better suited.

Parallel robots have been found in many applications where the work requirements are beyond the capabilities of serial robots. For example, mouth movements of chewing foods can be generated by a parallel robot. In this paper, the issue of dynamic position and force control of a chewing robot with a 6RSS mechanism is addressed. The kinematic and dynamic models of a generic 6RSS robot are developed and are then simplified considering the special features of a practical chewing robot and the requirements of controller design. An impedance control scheme is proposed to achieve the position and force control of the robot. A detailed description on the steps to implement the controller is also presented.

Object tracking is an important process for many applications in computer vision. This process must be implemented in a discrete manner because the images are available only at certain periods. A discrete integral sliding mode algorithm is proposed to control a stereo vision system and perform the aforementioned second task. The kinematic model of the structure is obtained using geometric algebra. The performance of the controller is compared with proportional-integral-derivative control via simulation. The implementation for a pan tilt unit is presented in real time. The algorithm presents a good and robust performance.