No abstract

Abstract-This position paper proposes that the study of embodied cognitive agents, such as humanoid robots, can advance our understanding of the cognitive development of complex sensorimotor, linguistic, and social learning skills. This in turn will benefit the design of cognitive robots capable of learning to handle and manipulate objects and tools autonomously, to cooperate and communicate with other robots and humans, and to adapt their abilities to changing internal, environmental, and social conditions. Four key areas of research challenges are discussed, specifically for the issues related to the understanding of: 1) how agents learn and represent compositional actions; 2) how agents learn and represent compositional lexica; 3) the dynamics of social interaction and learning; and 4) how compositional action and language representations are integrated to bootstrap the cognitive system. The review of specific issues and progress in these areas is then translated into a practical roadmap based on a series of milestones. These milestones provide a possible set of cognitive robotics goals and test scenarios, thus acting as a research roadmap for future work on cognitive developmental robotics.

Programming robots to carry out useful tasks is both a complex and non-trivial exercise. A simple and intuitive method to allow humans to train and shape robot behaviour is clearly a key goal in making this task easier. This paper describes an approach to this problem based on studies of social animals where two teaching strategies are applied to allow a human teacher to train a robot by moulding its actions within a carefully scaffolded environment. Within these enviroments sets of competences can be built by building state/action memory maps of the robot's interaction within that environment. These memory maps are then polled using a k-nearest neighbour based algorithm to provide a generalised competence. We take a novel approach in building the memory models by allowing the human teacher to construct them in a hierarchical manner. This mechanism allows a human trainer to build and extend an action-selection mechanism into which new skills can be added to the robot's repertoire of existing competencies. These techniques are implemented on physical Khepera miniature robots and validated on a variety of tasks.

Care issues and costs associated with an increasing elderly population are becoming a major concern for many countries. The use of assistive robots in "smart-home" environments has been suggested as a possible partial solution to these concerns. A challenge is the personalization of the robot to meet the changing needs of the elderly person over time. One approach is to allow the elderly person, or their carers or relatives, to make the robot learn activities in the smart home and teach it to carry out behaviors in response to these activities. The overriding premise being that such teaching is both intuitive and "nontechnical." To evaluate these issues, a commercially available autonomous robot has been deployed in a fully sensorized but otherwise ordinary suburban house. We describe the design approach to the teaching, learning, robot, and smart home systems as an integrated unit and present results from an evaluation of the teaching component with 20 participants and a preliminary evaluation of the learning component with three participants in a human-robot interaction experiment. Participants reported findings using a system usability scale and ad-hoc Likert questionnaires. Results indicated that participants thought that this approach to robot personalization was easy to use, useful, and that they would be capable of using it in real-life situations both for themselves and for others.

It is essential for robots working in close proximity to people to be both safe and trustworthy. We present a case study on formal verification for a high-level planner/scheduler for the Care-O-bot, an autonomous personal robotic assistant. We describe how a model of the Care-O-bot and its environment was developed using Brahms, a multiagent workflow language. Formal verification was then carried out by automatically translating this model to the input language of an existing model checker. Four sample properties based on system requirements were verified. We then refined the environment model three times to increase its accuracy and the persuasiveness of the formal verification results. The first refinement uses a user activity log based on real-life experiments, but is deterministic. The second refinement uses the activities from the user activity log nondeterministically. The third refinement uses "conjoined activities" based on an observation that many user activities can overlap. The four samples properties were verified for each refinement of the environment model. Finally, we discuss the approach of environment model refinement with respect to this case study.

We report on experiments and analyses dealing with the acquisition of lexical meaning in which prosodic analysis and extraction of salient words are associated with a robots sensorimotor perceptions in an attempt to ground these words in the robots own embodied sensorimotor experience. We focus here on three key areas, the selection of salient words based on prosodic clues, expression of words by the robot at a two-word stage to reflect learning and grammatically correct presentation, and an in-depth analysis of the relationship between words and the robots sensorimotor perceptions.

Modern theories on early child language acquisition tend to focus on referential words, mostly nouns, labeling concrete objects, or physical properties. In this experimental proof-ofconcept study, we show how nonreferential negation words, typically belonging to a child's first ten words, may be acquired. A childlike humanoid robot is deployed in speech-wise unconstrained interaction with naïve human participants. In agreement with psycholinguistic observations, we corroborate the hypothesis that affect plays a pivotal role in the socially distributed acquisition process where the adept conversation partner provides linguistic interpretations of the affective displays of the less adept speaker. Negation words are prosodically salient within intent interpretations that are triggered by the learner's display of affect. From there they can be picked up and used by the budding language learner which may involve the grounding of these words in the very affective states that triggered them in the first place. The pragmatic analysis of the robot's linguistic performance indicates that the correct timing of negative utterances is essential for the listener to infer the meaning of otherwise ambiguous negative utterances. In order to assess the robot's performance thoroughly comparative data from psycholinguistic studies of parent-child dyads is needed highlighting the need for further interdisciplinary work.

Imitative learning and learning by observation are social mechanisms that allow a robot to acquire knowledge from a human or another robot. However to be able to obtain skills in this way the robot faces many complex issues, one of which is that of finding solutions to the correspondence problem. Evolutionary predecessors to observational imitation may have been self-imitation where an agent avoids the complexities of the correspondence problem by learning and replicating actions it has experienced through the manipulation of its body. We investigate how a robotic control and teaching system using self-imitation can be constructed with reference to psychological models of motor control and ideas from social scaffolding seen in animals. Within these scaffolded environments sets of competencies can be built by constructing hierarchical state/action memory maps of the robot's interaction within that environment. The scaffolding process provides a mechanism to enable learning to be scaled up. The resulting system allows a human trainer to teach a robot new skills and modify skills that the robot may possess. Additionally the system allows the robot to notify the trainer when it is being taught skills it already has in its repertoire and to direct and focus its attention and sensor resources to relevant parts of the skill being executed. We argue that these mechanisms may be a first step towards the transformation from self-imitation to observational imitation. The system is validated on a physical pioneer robot that is taught using self-imitation to track, follow and point to a patterned object.