This paper proposes a framework for learning human-provided category labels that describe individual objects, pairwise object relationships, as well as groups of objects. The framework was evaluated using an experiment in which the robot interactively explored 36 objects that varied by color, weight, and contents. The proposed method allowed the robot not only to learn categories describing individual objects, but also to learn categories describing pairs and groups of objects with high recognition accuracy. Furthermore, by grounding the category representations in its own sensorimotor repertoire, the robot was able to estimate how similar two categories are in terms of the behaviors and sensory modalities that are used to recognize them. Finally, this grounded measure of similarity enabled the robot to boost its recognition performance when learning a new category by relating it to a set of familiar categories.

This paper proposes a theoretical model that enables a robot to partition its unlabeled sensorimotor experience with different objects into discrete clusters, each corresponding to a specific object. To solve this object individuation problem, the robot was trained to detect whether two perceptual stimuli were produced by the same object or by two different objects. The model was tested using a large-scale experiment in which a humanoid robot explored 100 different objects by performing a variety of exploratory behaviors on them and detecting the resulting sensory feedback from several sensory modalities. The results show that with a small amount of prior training, the robot's model was able to successfully individuate the objects with a high degree of accuracy.

This paper describes a set of experiments in which an upper-torso humanoid robot learned to slide a card through a card reader. The small size and the flexibility of the card presented a number of manipulation challenges for the robot. First, because most of the card is occluded by the card reader and the robot's hand during the sliding process, visual feedback is useless for this task. Second, because the card bends easily, it is difficult to distinguish between bending and hitting an obstacle in order to correct the sliding trajectory. To solve these manipulation challenges this paper proposes a method for constraint detection that uses only proprioceptive data. The method uses dynamic joint torque thresholds that are calibrated using the robot's movements in free space. The experimental results show that using this method, the robot can detect when the movement of the card is constrained and modify the sliding trajectory in real time, which makes solving this task possible.

This paper describes an approach to solving insertion tasks by a robot that uses exploratory behaviors and proprioceptive feedback. The approach was inspired by the developmental progression of insertion abilities in both chimpanzees and humans (Hayashi et al. 2006). Before mastering insertions, the infants of the two species undergo a stage where they only press objects against other objects without releasing them. Our goal was to emulate this developmental stage on a robot to see if it may lead to simpler representations for insertion tasks. Experiments were performed using a shapesorter puzzle with three different blocks and holes.

The ability to form and recognize object categories is fundamental to human intelligence. This paper proposes a behavior-grounded relational classification model that allows a robot to recognize the categories of household objects. In the proposed approach, the robot initially explores the objects by applying five exploratory behaviors (lift, shake, drop, crush and push) on them while recording the proprioceptive and auditory sensory feedback produced by each interaction. The sensorimotor data is used to estimate multiple measures of similarity between the objects, each corresponding to a specific coupling between an exploratory behavior and a sensory modality. A graph-based recognition model is trained by extracting features from the estimated similarity relations, allowing the robot to recognize the category memberships of a novel object based on the object's similarity to the set of familiar objects. The framework was evaluated on an upper-torso humanoid robot with two large sets of household objects. The results show that the robot's model is able to recognize complex object categories (e.g., metal objects, empty bottles, etc.) significantly better than chance.

This paper proposes a method for interactive surface recognition and surface categorization by a humanoid robot using a vibrotactile sensory modality. The robot was equipped with an artificial fingernail that had a built-in three-axis accelerometer. The robot interacted with 20 different surfaces by performing five different exploratory scratching behaviors on them. Surface-recognition models were learned by coupling frequency-domain analysis of the vibrations detected by the accelerometer with machine learning algorithms, such as support vector machine (SVM) and k-nearest neighbors (k -NN). The results show that by applying several different scratching behaviors on a test surface, the robot can recognize surfaces better than with any single behavior alone. The robot was also able to estimate a measure of similarity between any two surfaces, which was used to construct a grounded hierarchical surface categorization.

The ability to form object categories is an important milestone in human infant development. We propose a framework that allows a robot to form a unified object categorization from several interactions with objects. This framework is consistent with the principle that robot learning should be ultimately grounded in the robot's perceptual and behavioral repertoire. This paper builds upon our previous work by adding more exploratory behaviors (now 6 instead of 1) and by employing consensus clustering for finding a single, unified object categorization. The framework was tested on a container/non-container categorization task with 20 objects.

Active object exploration is one of the hallmarks of human and animal intelligence. Research in psychology has shown that the use of multiple exploratory behaviors is crucial for learning about objects. Inspired by such research, recent work in robotics has demonstrated that by performing multiple exploratory behaviors a robot can dramatically improve its object recognition rate. But what is the cause of this improvement? To answer this question, this paper examines the conditions under which combining information from multiple behaviors and sensory modalities leads to better object recognition results. Two different problems are considered: interactive object recognition using auditory and proprioceptive feedback, and surface texture recognition using tactile and proprioceptive feedback. Analysis of the results shows that metrics designed to estimate classifier model diversity can explain the improvement in recognition accuracy. This finding establishes, for the first time, an important link between empirical studies of exploratory behaviors in robotics and theoretical results on boosting in machine learning.

To function in human-inhabited environments a robot must be able to press buttons. There are literally thousands of different buttons, which produce various types of feedback when pressed. This work focuses on doorbell buttons, which provide auditory feedback. Our robot learned to predict if a specific pushing movement would press a doorbell button and produce a sound. The robot explored different buttons with random pushing behaviors and perceived the proprioceptive, tactile, and acoustic outcomes of these behaviors.

This paper describes an approach to interactive object categorization that couples exploratory behaviors and their resulting acoustic signatures to form object categories. The framework was tested with an upper-torso humanoid robot on a container/non-container categorization task. The robot used six exploratory behaviors (drop block, grasp, move, shake, flip, and drop object) and applied them to twenty objects. The results from this large-scale experimental study show that the robot was able to learn meaningful object categories using only acoustic information. The results also show that the quality of the categorization depends on the exploratory behavior used to derive it as some behaviors elicit more salient acoustic signatures than others.

Human beings can perceive object properties such as size, weight, and material type based solely on the sounds that the objects make when an action is performed on them. In order to be successful, the household robots of the near future must also be capable of learning and reasoning about the acoustic properties of everyday objects. Such an ability would allow a robot to detect and classify various interactions with objects that occur outside of the robot's field of view. This paper presents a framework that allows a robot to infer the object and the type of behavioral interaction performed with it from the sounds generated by the object during the interaction. The framework is evaluated on a 7-d.o.f. Barrett WAM robot which performs grasping, shaking, dropping, pushing and tapping behaviors on 36 different household objects. The results show that the robot can learn models that can be used to recognize objects (and behaviors performed on objects) from the sounds generated during the interaction. In addition, the robot can use the learned models to estimate the similarity between two objects in terms of their acoustic properties.

This paper introduces a novel approach to representing and learning tool affordances by a robot. The tool representation described here uses a behavior-based approach to ground the tool affordances in the behavioral repertoire of the robot. The representation is learned during a behavioral babbling stage in which the robot randomly chooses different exploratory behaviors, applies them to the tool, and observes their effects on environmental objects. The paper shows how the autonomously learned affordance representation can be used to solve tool-using tasks by dynamically sequencing the exploratory behaviors based on their expected outcomes. The quality of the learned representation was tested on extension-of-reach tool-using tasks.

Q-learning has often been used to learn primitive behaviors, or to coordinate a limited set of motor skills. However, the complexity of the algorithm increases exponentially with the number of states the robot can be in and the number of actions that it can take. Therefore, it is natural to try to reduce the number of states and actions in order to improve the efficiency of the algorithm. Robot behaviors and behavioral assemblages provide a good level of abstraction which could be used to speed up robot learning. Instead of coordinating a set of primitives, we use Q-learning to coordinate a set of well tested behavioral assemblages to accomplish a robot mission. The domain for our experiments is a simple intercept mission. This paper also explores the effects of imperfect perceptual algorithms on learning when this approach is used.