Çetin Meriçli

Market-Driven Multi-Agent Collaboration in Robot Soccer Domain

In recent robotic studies, in many key areas decoupled multi-agent systems have become more popular than complex single agent systems, where the former is more robust, fast and cheap to implement. The most important reason behind this preference is to eliminate the possibility of single point of failure, which is a vital concern for single complex agents. Usage of decoupled multi-agent systems may also reduce the total cost of the entire system when it is possible to use a team of single cheap robots for performing complex tasks, instead of building up a single complex and expensive robot to satisfy all the system needs. As a result, typically when a team of robots is used, the system throughput increases while the total cost decreases. Since the robots usually have simpler physical structures, generally less complicated controller programs are necessary to manipulate the agents. The decoupled behaviors of agents can cause communication and coordination problems, however. The studies in (Dudek

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.

CoBots: Collaborative robots servicing multi-floor buildings

In this video we briefly illustrate the progress and contributions made with our mobile, indoor, service robots CoBots (Collaborative Robots), since their creation in 2009. Many researchers, present authors included, aim for autonomous mobile robots that robustly perform service tasks for humans in our indoor environments. The efforts towards this goal have been numerous and successful, and we build upon them. However, there are clearly many research challenges remaining until we can experience intelligent mobile robots that are fully functional and capable in our human environments.

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.

Multi-humanoid world modeling in Standard Platform robot soccer

In the RoboCup Standard Platform League (SPL), the robot platform is the same humanoid NAO robot for all the competing teams. The NAO humanoids are fully autonomous with two onboard directional cameras, computation, multi-joint body, and wireless communication among them. One of the main opportunities of having a team of robots is to have robots share information and coordinate. We address the problem of each humanoid building a model of the world in real-time, given a combination of its own limited sensing, known models of actuation, and the communicated information from its teammates. Such multi-humanoid world modeling is challenging due to the biped motion, the limited perception, and the tight coupling between behaviors, sensing, localization, and communication. We describe the real-world opportunities, constraints and limitations imposed by the NAO humanoid robots. We contribute a modeling approach that differentiates among the motion model of different objects, in terms of their dynamics, namely the static landmarks (e.g., goal posts, lines, corners), the passive moving ball, and the controlled moving robots, both teammates and adversaries. We present experimental results with the NAO humanoid robots to illustrate the impact of our multi-humanoid world modeling approach. The challenges and approaches we present are relevant to the general problem of assessing and sharing information among multiple humanoid robots acting in a world with multiple types of objects.

Fast human detection for indoor mobile robots using depth images

A human detection algorithm running on an indoor mobile robot has to address challenges including occlusions due to cluttered environments, changing backgrounds due to the robots motion, and limited on-board computational resources. We introduce a fast human detection algorithm for mobile robots equipped with depth cameras. First, we segment the raw depth image using a graph-based segmentation algorithm. Next, we apply a set of parameterized heuristics to filter and merge the segmented regions to obtain a set of candidates. Finally, we compute a Histogram of Oriented Depth (HOD) descriptor for each candidate, and test for human presence with a linear SVM. We experimentally evaluate our approach on a publicly available dataset of humans in an open area as well as our own dataset of humans in a cluttered cafe environment. Our algorithm performs comparably well on a single CPU core against another HOD-based algorithm that runs on a GPU even when the number of training examples is decreased by half. We discuss the impact of the number of training examples on performance, and demonstrate that our approach is able to detect humans in different postures (e.g. standing, walking, sitting) and with occlusions.

Task Refinement for Autonomous Robots using Complementary Corrective Human Feedback

A robot can perform a given task through a policy that maps its sensed state to appropriate actions. We assume that a hand-coded controller can achieve such a mapping only for the basic cases of the task. Refining the controller becomes harder and gets more tedious and error prone as the complexity of the task increases. In this paper, we present a new learning from demonstration approach to improve the robots performance through the use of corrective human feedback as a complement to an existing hand-coded algorithm. The human teacher observes the robot as it performs the task using the hand-coded algorithm and takes over the control to correct the behavior when the robot selects a wrong action to be executed. Corrections are captured as new state-action pairs and the default controller output is replaced by the demonstrated corrections during autonomous execution when the current state of the robot is decided to be similar to a previously corrected state in the correction database. The proposed approach is applied to a complex ball dribbling task performed against stationary defender robots in a robot soccer scenario, where physical Aldebaran Nao humanoid robots are used. The results of our experiments show an improvement in the robots performance when the default hand-coded controller is augmented with corrective human demonstration.

All Bids for One and One Does for All: Market-Driven Multi-agent Collaboration in Robot Soccer Domain

In this paper, a novel market-driven collaborative task allocation algorithm called “Collaboration by competition / cooperation” for the robot soccer domain is proposed and implemented. In robot soccer, two teams of robots compete with each other to win the match. For the benefit of the team, the robots should work collaboratively, whenever possible. The market-driven approach applies the basic properties of free market economy to a team of robots for increasing the profit of the team as much as possible. The experimental results show that the approach is robust and flexible and the developed team is more succcessful than its opponents.

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.

Multi-resolution Corrective Demonstration for Efficient Task Execution and Refinement

Computationally efficient task execution is very important for autonomous mobile robots endowed with limited on-board computational resources. Most robot control approaches assume a fixed state and action representation, and use a single algorithm to map states to actions. However, not all situations in a given task require equally complex algorithms and equally detailed state and action representations. The main motivation for this work is a desire to reduce the computational footprint of performing a task by allowing the robot to run simpler algorithms whenever possible, and resort to a more complex algorithm only when needed. We contribute the Multi-Resolution Task Execution (MRTE) algorithm that utilizes human feedback to learn a mapping from a given state to an appropriate detail resolution consisting of a state and action representation, and an algorithm providing a mapping from states to actions at that resolution. The robot learns a policy from human demonstration to switch between different detail resolutions as needed while favoring lower detail resolutions to reduce computational cost of task execution. We then present the Model Plus Correction (M+C) algorithm to improve the performance of an algorithm using corrective human feedback without modifying the algorithm itself. Finally, we introduce the Multi-Resolution Model Plus Correction (MRM+C) algorithm as a combination of MRTE and M+C. MRM+C learns how to select an appropriate detail resolution to operate at in a given state from human demonstration. Furthermore, it allows the teacher to provide corrective demonstration at different detail resolutions to improve overall task execution performance. We provide formal definitions of MRTE, M+C, and MRM+C algorithms, and show how they relate to general robot control problem and Learning from Demonstration (LfD) approach. We present experimental results de-monstrating the effectiveness of proposed methods on a goal-directed humanoid obstacle avoidance task.