International Journal of Humanoid Robotics

This paper presents a novel combination of visual servoing (VS) control and neural network (NN) learning on humanoid dual-arm robot. A VS control system is built by using stereo vision to obtain the 3D point cloud of a target object. A least square-based method is proposed to reduce the stochastic error in workspace calibration. An NN controller is designed to compensate for the effect of uncertainties in payload and other parameters (both internal and external) during the tracking control. In contrast to the conventional NN controller, a deterministic learning technique is utilized in this work, to enable the learned neural knowledge to be reused before current dynamics changes. A skill transfer mechanism is also developed to apply the neural learned knowledge from one arm to the other, to increase the neural learning efficiency. Tracked trajectory of object is used to provide target position to the coordinated dual arms of a Baxter robot in the experimental study. Robotic implementations has demonstrated the efficiency of the developed VS control system and has verified the effectiveness of the proposed NN controller with knowledge-reuse and skill transfer features.

Gesture recognition plays an important role in human–computer interaction. However, most existing methods are complex and time-consuming, which limit the use of gesture recognition in real-time environments. In this paper, we propose a static gesture recognition system that combines depth information and skeleton data to classify gestures. Through feature fusion, hand digit gestures of 0–9 can be recognized accurately and efficiently. According to the experimental results, the proposed gesture recognition system is effective and robust, which is invariant to complex background, illumination changes, reversal, structural distortion, rotation, etc. We have tested the system both online and offline which proved that our system is satisfactory to real-time requirements, and therefore it can be applied to gesture recognition in real-world human–computer interaction systems.

This paper proposes a precise driving control algorithm by utilizing the slip pattern of an inverted ball-driving robot with Mecanum wheels. The slip pattern of the ball-driving robot has an eight-leaves flower shape, which can be utilized to plan the driving direction of the ball-robot and to minimize the slip. When there exists the slip, the localization of the ball-robot becomes very coarse because of the uncertainties caused by the slip. By minimizing the slip using the analyzed slip pattern, the precise driving of the inverted ball-driving robot with Mecanum wheels can be achieved. Real experiments with a self-manufactured ball-robot have been carried out to show the preciseness of the driving based upon the proposed algorithm compared to the conventional algorithms.

This work presents the kinematic and dynamic modeling of a human–wheelchair system which considers that its center of mass is not located in the middle of the wheel’s axle. Furthermore, a novel motion controller is presented for a human–wheelchair system, which is capable of performing positioning and path-following tasks in human-shared environments. This controller design is based on two cascaded subsystems: a kinematic controller, and a dynamic controller that compensates the dynamics of the human–wheelchair system. Additionally, an algorithm based on fuzzy-logic is proposed and incorporated in the aforementioned path-following control for pedestrian collision avoidance. This methodology considers to quantify heuristics social rules to make a balance between modulating velocity or direction during the avoidance. Three different interference cases, commonly found during walking events, are tested in a structured scenario. The experimental results demonstrate that the system is capable of overcoming many usual interference situations with human obstacles. A good performance of the path-following control is also verified.

The existing procedures for autism spectrum disorder (ASD) diagnosis are often time consuming and tiresome both for highly-trained human evaluators and children, which may be alleviated by using humanoid robots in the diagnostic process. Hence, this paper proposes a framework for robot-assisted ASD evaluation based on partially observable Markov decision process (POMDP) modeling, specifically POMDPs with mixed observability (MOMDPs). POMDP is broadly used for modeling optimal sequential decision making tasks under uncertainty. Spurred by the widely accepted autism diagnostic observation schedule (ADOS), we emulate ADOS through four tasks, whose models incorporate observations of multiple social cues such as eye contact, gestures and utterances. Relying only on those observations, the robot provides an assessment of the child’s ASD-relevant functioning level (which is partially observable) within a particular task and provides human evaluators with readable information by partitioning its belief space. Finally, we evaluate the proposed MOMDP task models and demonstrate that chaining the tasks provides fine-grained outcome quantification, which could also increase the appeal of robot-assisted diagnostic protocols in the future.