Autism is a developmental condition that can cause significant social, communication, and behavioral challenges. Children on the autism spectrum may have difficulties developing verbal communication skills, understanding what others say, or communicating through non-verbal cues. Similar difficulties are experienced by children with developmental delay. A recent trend in robotics is the design and implementation of robots to assist during therapy and education of children with learning difficulties. Although encouraging results suggests that robots can be beneficial, there has been limited work on the long-term impact of these tools on the verbal communication skills of children with autism or developmental delay. This paper explores the impact of robots on the verbal communication skills of secondary aged students with moderate to severe intellectual disabilities and autism. A qualitative study was carried out, via focus groups and interviews with parents, carers and staff members, 24 months after the introduction of two humanoid robots into the disability unit of a public secondary school. Results show that humanoid robots can provide benefits in articulation, verbal participation and spontaneous conversation in these young adults. Three exemplars are presented.

For some time now, researchers have explored the use of social robots as tools to assist during therapy and education for children with intellectual disabilities and autism. Although encouraging results suggest that robots can be beneficial, there has been minimal progress in integrating this technology as formal tools to supplement therapy and education. A significant reason behind this is that the benefits of using robots have been demonstrated primarily in short-term case studies, with very few demonstrations over large samples. This study explores the impact of a prolonged exposure to socially-assistive robots in an education context with secondary students with intellectual disabilities and autism. The evaluation was carried out 24 months after the introduction of two different robots into the disability unit from a public secondary school. Participants responded positively toward the use of robots at the school, and would recommend other schools to use them as well. Benefits, challenges, and limitations of the robots are discussed.

This paper describes the continued development of a system to provide early assessment of speech development issues in children and better triaging to professional services. Whilst corpora of children's speech are increasingly available, recognition of disordered children's speech is still a data-scarce task. Transfer learning methods have been shown to be effective at leveraging out-of-domain data to improve ASR performance in similar data-scarce applications. This paper combines transfer learning, with previously developed methods for constrained decoding based on expert speech pathology knowledge and knowledge of the target text. Results of this study show that transfer learning with out-of-domain adult speech can improve phoneme recognition for disordered children's speech. Specifically, a Deep Neural Network (DNN) trained on adult speech and fine-tuned on a corpus of disordered children's speech reduced the phoneme error rate (PER) of a DNN trained on a children's corpus from 16.3% to 14.2%. Furthermore, this fine-tuned DNN also improved the performance of a Hierarchal Neural Network based acoustic model previously used by the system with a PER of 19.3%. We close with a discussion of our planned future developments of the system.

Experiments were conducted to quantify the direction of robot approach that a pair of seated people find most comfortable. Three maximally-different seating configurations and eight directions of robot approach were considered. The data were analysed using Rayleigh's test of uniformity, a directional statistics method. Results from 140 participants showed robot approach directions that minimise participant discomfort align spatially with regions that allow good sight of the robot by both people and are centred on the pair's largest unoccupied area of p-space.

Latest research in robotics is driven by a desire to develop systems that would allow humans to interact with robots in natural and intuitive ways. In this paper we investigate peoples' initial believes, expectations and ascribed mental capabilities of a robot when they encounter it for the first time and how these change after interacting with the robot. The differences between two robots, which differ in morphology and behavior, are compared.

This study focuses on differences and similarities of perception of a small humanoid robot between Japanese and Australian participants. Two conditions were investigated: participants actively interacting with the robot and bystanders observing the interaction. Experimental results suggested that, while the robot was perceived as highly likeable, Japanese participants rated the robot higher for animacy, intelligence and safety. Furthermore, passive observations of the interaction (rather than active interaction) resulted in higher ratings by Japanese participants for anthropomorphism, animacy, intelligence and safety. The findings are discussed in terms of cultural background and robot perception.

This paper explores the uncanny valley in a possible future when humans and androids are indistinguishable from each other. In this setting, humans will by definition experience a condition of radical uncertainty. We investigate the strategies people use to attempt to distinguish between human-like entities, both humans and androids. Experimental results demonstrate that visual and contextual information are important during entity identification, but under conditions of radical uncertainty, a feeling of discomfort and rejection towards human-like entities - both humans and androids - may be experienced.

The study of human-robot interaction commonly occurs in a laboratory environment in an endeavour to control the experimental conditions. Although convenient for the experimenters, a laboratory setting can constrain and therefore alter a participant's behaviour. In this paper we introduce the method of `open experimentation' in which human-robot interaction studies take place outside the laboratory in a public social environment. Results from over 4,300 participants interacting with a humanoid robot, Diamandini, demonstrate that the general population preferred to observe the robot from a social distance, while children were comfortable to interact more closely with it and for extended periods of time.

During social interaction, humans extract important information from tactile stimuli that improves their understanding of the interaction. The development of a similar capacity in a robot will contribute to the future success of intuitive human-robot interaction. This paper presents a method of touch sensing based on the principle of electrical impedance tomography (EIT) that can be used to implement a large, flexible and stretchable artificial sensitive skin for robots. A classifier based on the “LogitBoost” algorithm is used to classify the modality of six different types of touch on an experimental EIT-based skin. Experiments showed that the modality of touch was correctly classified in approximately 80% of the trials. This is comparable with the experimental accuracy of a human touch recipient. The classification accuracies show significant improvements from previous classification algorithms applied to different artificial sensitive skins.

Electrical impedance tomography (EIT) is a technique used to estimate the internal conductivity of an electrically conductive body by using measurements made only at its boundary. If this body is made of a thin, flexible, and stretchable material that responds to touch with local changes in conductivity, it can be used to create an artificial sensitive skin. Mathematically, the EIT reconstruction problem is an ill-posed nonlinear inverse problem in which it is commonly assumed that electrodes are located only on the surface of the body. In a thin sensitive skin, however, electrodes can readily be located within the 2-D conducting domain. This paper compares existing electrode-drive patterns with new patterns in which a number of reference electrodes are located inside of the sensitive skin. Simulation results and experimental data show improvements in both resolution and robustness to noise of the reconstructed image. These improvements are shown to be consistent for several commonly used regularization methods.