Tekin Meriçli

Joint Attention by Gaze Interpolation and Saliency

Joint attention, which is the ability of coordination of a common point of reference with the communicating party, emerges as a key factor in various interaction scenarios. This paper presents an image-based method for establishing joint attention between an experimenter and a robot. The precise analysis of the experimenters eye region requires stability and high-resolution image acquisition, which is not always available. We investigate regression-based interpolation of the gaze direction from the head pose of the experimenter, which is easier to track. Gaussian process regression and neural networks are contrasted to interpolate the gaze direction. Then, we combine gaze interpolation with image-based saliency to improve the target point estimates and test three different saliency schemes. We demonstrate the proposed method on a human-robot interaction scenario. Cross-subject evaluations, as well as experiments under adverse conditions (such as dimmed or artificial illumination or motion blur), show that our method generalizes well and achieves rapid gaze estimation for establishing joint attention.

Joint visual attention modeling for naturally interacting robotic agents

This paper elaborates on mechanisms for establishing visual joint attention for the design of robotic agents that learn through natural interfaces, following a developmental trajectory not unlike infants. We describe first the evolution of cognitive skills in infants and then the adaptation of cognitive development patterns in robotic design. A comprehensive outlook for cognitively inspired robotic design schemes pertaining to joint attention is presented for the last decade, with particular emphasis on practical implementation issues. A novel cognitively inspired joint attention fixation mechanism is defined for robotic agents.

Push-manipulation of complex passive mobile objects using experimentally acquired motion models

In a realistic mobile push-manipulation scenario, it becomes non-trivial and infeasible to build analytical models that will capture the complexity of the interactions between the environment, each of the objects, and the robot as the variety of objects to be manipulated increases. We present an experience-based push-manipulation approach that enables the robot to acquire experimental models regarding how pushable real world objects with complex 3D structures move in response to various pushing actions. These experimentally acquired models can then be used either (1) for trying to track a collision-free guideline path generated for the object by reiterating pushing actions that result in the best locally-matching object trajectories until the goal is reached, or (2) as building blocks for constructing achievable push plans via a Rapidly-exploring Random Trees variant planning algorithm we contribute and executing them by reiterating the corresponding trajectories. We extensively experiment with these two methods in a 3D simulation environment and demonstrate the superiority of the achievable planning and execution concept through safe and successful push-manipulation of a variety of passively mobile pushable objects. Additionally, our preliminary tests in a real world scenario, where the robot is asked to arrange a set of chairs around a table through achievable push-manipulation, also show promising results despite the increased perception and action uncertainty, and verify the validity of our contributed method.

Practical extensions to vision-based monte carlo localization methods for robot soccer domain

This paper proposes a set of practical extensions to the vision-based Monte Carlo localization (MCL) for RoboCup Sony AIBO legged robot soccer domain. The main disadvantage of AIBO robots is that they have a narrow field of view so the number of landmarks seen in one frame is usually not enough for geometric calculation. MCL methods have been shown to be accurate and robust in legged robot soccer domain but there are some practical issues that should be handled in order to maintain stability/elasticity ratio in a reasonable level. In this work, we presented four practical extensions in which two of them are novel approaches and the remaining ones are different from the previous implementations.

Introduction to Autonomous Mobile Robotics Using "Lego Mindstorms" NXT.

Teaching the fundamentals of robotics to computer science undergraduates requires designing a well-balanced curriculum that is complemented with hands-on applications on a platform that allows rapid construction of complex robots, and implementation of sophisticated algorithms. This paper describes such an elective introductory course where the Lego Mindstorms NXT kits are used as the robot platform. The aims, scope and contents of the course are presented, and the design of the laboratory sessions as well as the term projects, which address several core problems of robotics and artificial intelligence simultaneously, are explained in detail.

Soccer without intelligence

Robot soccer is an excellent testbed to explore innovative ideas and test the algorithms in multi-agent systems (MAS) research. A soccer team should play in an organized manner in order to score more goals than the opponent, which requires well-developed individual and collaborative skills, such as dribbling the ball, positioning, and passing. However, none of these skills needs to be perfect and they do not require highly complicated models to give satisfactory results. This paper proposes an approach inspired from ants, which are modeled as Braitenberg vehicles for implementing those skills as combinations of very primitive behaviors without using explicit communication and role assignment mechanisms, and applying reinforcement learning to construct the optimal state-action mapping. Experiments demonstrate that a team of robots can indeed learn to play soccer reasonably well without using complex environment models and state representations. After very short training sessions, the team started scoring more than its opponents that use complex behavior codes, and as a result of having very simple state representation, the team could adapt to the strategies of the opponent teams during the games.

A Case-Based Approach to Mobile Push-Manipulation

The complexity of the potential physical interactions between the robot, each of the pushable objects, and the environment is vast in realistic mobile push-manipulation scenarios. This makes it non-trivial to write generic analytical motion and interaction models or tune the parameters of physics engines for every robot, object, and environment combination. We present a case-based approach to push-manipulation that allows the robot to acquire, through interaction and observation, a set of discrete, experimental, probabilistic motion models (i.e. probabilistic cases) for pushable passively-mobile real world objects. These probabilistic cases are then used as building blocks for constructing achievable push plans to navigate the object of interest to the desired goal pose as well as to potentially push the movable obstacles out of the way in cluttered task environments. Additionally, incremental acquisition and updating of the probabilistic cases allows the robot to adapt to the changes in the environment, such as increased mass due to loading of the object of interest for transportation purposes. The purely interaction and observation driven nature of our method makes it robot, object, and environment (real or simulated) independent, as we demonstrate through validation tests in a real world setup in addition to extensive experimentation in simulation.

Let's be honest: A controlled field study of ethical behavior in the presence of a robot

Human-robot collaboration will increasingly take place in human social settings, including contexts where ethical and honest behavior is paramount. How might these robots affect human honesty? In this paper, we present first evidence of how a robots presence affects peoples ethical behavior in a controlled field study. We observed people passing by a food plate marked as “reserved”, comparing three conditions: no observer, a human observer, and a robot observer. We found that a human observer elicits less attention than a robot, but evokes more of a socially normative presence causing people to act honestly. Conversely, we found that a robot observer elicits more attention, engagement, and a monitoring presence. But even though people were suspicious that they were being monitored, they still behaved dishonestly in the robot observer condition.

A Robust Statistical Collision Detection Framework for Quadruped Robots

In order to achieve its tasks in an effective manner, an autonomous mobile robot must be able to detect and quickly recover from collisions. This paper proposes a new solution to the problem of detecting collisions during omnidirectional motion of a quadruped robot equipped with an internal accelerometer. We consider this as an instance of general signal processing and statistical anomaly detection problems. We find that temporal accelerometer readings examined in the frequency domain are good indicators of regularities (normal motion) and novel situations (collisions). In the course of time, the robot builds a probabilistic model that captures its proprioceptive properties while walking without obstruction and uses that model to determine whether there is an abnormality in the case of an unfamiliar pattern. The approach does not depend on walk characteristics and the walking algorithm used, and is insensitive to the surface texture that the robot walks on as long as the surface is flat. The experiments demonstrate quite fast and successful detection of collisions independent of the point of contact with an acceptably low false positive rate.

Robot detection with a cascade of boosted classifiers based on haar-like features

Accurate world modeling is important for efficient multi-robot planning in robot soccer. Visual detection of the robots on the field in addition to all other objects of interest is crucial to achieve this goal. The problem of robot detection gets even harder when robots with only on board sensing capabilities, limited field of view, and restricted processing power are used. This work extends the real-time object detection framework proposed by Viola and Jones, and utilizes the unique chest and head patterns of Nao humanoid robots to detect them in the image. Experiments demonstrate rapid detection with an acceptably low false positive rate, which makes the method applicable for real-time use.