Research
- The learning and use of affordances in robots:
My research in cognitive robotics has started with
the investigation of the notion of affordances
within the MACS project. The notion was introduced
by J. J. Gibson to explain how inherent “values” and
“meanings” of things in the environment can be
directly perceived and how this information can be
linked to the action possibilities offered to the
organism by the environment. Although introduced in
psychology, the concept influenced studies in other
fields ranging from human–computer interaction to
autonomous robotics. Dr. We reviewed the use of the
term in different fields, with particular emphasis
on its use in autonomous robotics and proposed a
formalization of affordances towards using it at
different levels of robot control ranging from
perceptual learning to planning.
Through implementing parts of the formalism, it was shown that a mobile robot can learn to perceive the traversability affordances after interacting with its environment. The results demonstrated that the robot was able to achieve perceptual economy after learning and that the affordances generalized well in the real world. In our subsequent research, we used different robot manipulators (including the iCub humanoid robot) to learn the manipulative affordances, such as push-ability, reach-ability and grasp-ability, of objects. We have shown that, the affordances that were learned allowed the robot to do goal emulation and planning in its perceptual space.
- The grounding and use of verbs and nouns on the
iCub humanoid robot: This line of research
focused on how concepts represented by verbs and
nouns in language can be grounded in the
sensory-motor experiences of the robot, and was
funded initially by the ROSSI Project. Later, I was
awarded one of the free iCub humanoid platforms
(iCubROSSI), as well as a complementary TUBITAK
project (TOOL) towards carrying out his research.
Specifically, we utilized the notion of affordances to argue that verbs tend to refer to the generation of a specific type of effect rather than a specific type of action. Then, we show how a robot can form these concepts through interactions with the environment and how humans can use these concepts to ease their communication with the robots. We also took a novel approach to nouns that specify different classes of object, by arguing that novel objects should be classified not through their visual looks (as is usually done in object recognition studies in computer vision) but through the affordances it offers to the robot.
We demonstrated on the iCub humanoid robot platform that the concepts that the robot develops can be used to understand what a human performs, perform multi-step planning for reaching a goal state as well as to specify a goal to the robot using symbolic descriptions. In our final review demonstration, we have shown that the robot can respond successfully to commands such as “grasp ball”, by identifying a ball-like object (that is rollable) on its table and choosing the proper action (top-grasp instead of a side-grasp) to achieve its goal.
- Self-organized flocking in robots and UAVs:
This line of research, first funded by TUBITAK
through CRS Career Project and MQC Project funded by
Turkish Air Industries aimed at developing mobile
and aerial swarms with scalable coordination
algorithms for flocking. Within CRS, we developed a
mobile robot platform from scratch specifically for
swarm robotic studies and manufactured 20 of them.
The CD-sized robot had the ability to sense and
distinguish proximal objects and robots, as well as
the bearing of neighboring robots through some novel
sensing technology.
We proposed a coordination behavior that makes a swarm of mobile robots, initially connected via proximal sensing, be able to wander in an environment by moving as a coherent group in open space and to avoid obstacles as if it were a “super-organism”. Different from prior flocking studies, which relied on designated or elected leader within the group, or the use of a common goal/homing direction, our coordination behavior operated in a distributed manner with completely on-board sensing, making it the first successful self-organized flocking in literature.
Recently, Turkish Air Industries funded a project to develop a proof-of-concept system towards the coordinated flying of a group of quadcopters. Despite the lack of any prior experience, we were able to demonstrate the fully autonomous flying of three quadcopters in only 6 months. My research in this track is moving towards the development of on-board sensing/signaling systems to implement flocking.
Projects
- Multi-Quadcopter coordination. Funded by Turkish Air Industries. In this project, we developed a proof-of-concept system that demonstrated a group of three quadcopter platforms to fly autonomously (fly-by-click) in a desired formation. Despite the lack of experience on aerial platforms, we were able to demonstrate the system within only 6 months. Project duration: July 2011- February 2012. Funding: 20,000 Euro.
- Affordance-based concept formation and tool use in humanoid robots: Funded by TU¨ BI˙TAK (Turkish Scientific and Technical Council) as a 1001 project (No: 109E033). In this project, we use a 53 DOF iCub humanoid robot platform and its physics-based simulator to study 1) how concepts, such as nouns and verbs in language, can be created from affordances and how they can be used to communicate with humans, and (ii) how a humanoid robot can learn the affordances enabled to him by a set of tools, and can use them to achieve goals that are otherwise unreachable. Funding: 122,000 Euro. Duration: Sep 2009 - Sep 2012.
- Emergence of Communication in iCub through Sensorimotor and Social Interaction. This project proposal (Web: http://eris.liralab.it/wiki/Emergence of Communication in iCub through Sensorimotor and- Social Interaction) submitted to the Open Call (Web: http://eris.liralab.it/wiki/RobotCub Open Call) announced by the Robotcub (Web: http://www.robotcub.org) adlı project (funded within FP6 as an Integrated Project) was ranked 6th among 31 proposals and I was awarded (Evaluation report: http://eris.liralab.it/viki/images/a/a5/Opencall report.pdf ) a 53 DOF iCub humanoid robot platform (costing 255,000 Euro ). The robot, one of the most complex humanoid robot platforms available, is designed to be open and is adopted by more than 20 laboratories (located in 8 different countries mostly within EU) worldwide. More information about the robot platform is available at: http://icub.org
- ROSSI (Emergence of communication in RObots through Sensorimotor and Social Interaction): A StREP project funded in FP7 within the “Cognitive Systems, Interaction, Robotics” call of ICT.Within ROSSI, METU used the experimental results obtained from behavioral and brain imaging experiments on both affordances and language to develop computational models that allowed the iCub humanoid robot to develop concepts represented by verbs and nouns in the language and to recognize humans’ actions in real-time. Funding (METU share): 433,000 Euro . Duration: Mar 2008 - Mar 2011,Web: http://rossiproject.eu.
- MACS (Multi-sensory Autonomous Cognitive Systems Interacting with Dynamic Environments for Perceiving and Learning Affordances) Project: A StREP project funded by the EC in FP6 within the “Cognitive Systems” strategic objective call of IST. Within this project, we investigated how the concept of affordances, which was originally conceived in Psychology, can be used in autonomous robot control. We developed a formalization of this concept towards using it at different levels of robot control ranging from perceptual learning to planning and, developed and implemented a robot control architecture. Funding (METU share): 307,000 Euro. Duration: Sep 2004 - Jan 2008,Web: http://macs-eu.org.
- Controllable Robotic Swarms (Project no: 104E066): A Career project awarded by TUBITAK. The main scientific objective of the project is to investigate how and to what extend the dynamics of a robotic swarm can be externally controlled. Towards this end, we have successfully designed, developed and manufactured 20 CD-sized mobile robots specifically for conducting swarm robotics research. Using these robots, we achieved the first truly self-organized flocking in robot swarms and have shown that the direction of the flock can be steered externally by controlling only a subset of the flock. Funding: 82,000 Euro. Duration: Apr 2005 - Apr 2010.
- EUCOG II (2nd European Network for the Advancement of Artificial Cognitive Systems, Interaction and Robotics), Network of Excellence Project, FP7-ICT-EUCogII- 23128, Member, Member No: 1089, Feb 2009 - Jan 2012.
- EURON II (European Robotics Network), Network of Excellence Project, FP6-507728, Member, Member No: 137, METU representative (The first Turkish member of the network) , Sep 2003-Apr 2008.
- Sub-contractor for the SWARM-BOTS Project (Sep 2003 – Oct 2004): Participated as a sub-contractor for IRIDIA within the Swarm-bots project (described in detail later). Within this project, together with three students, developed PES (Parallelized Evolution System) as a platform that parallelizes the fitness evaluations of evolutionary methods over multiple computers connected via a network. METU budget: 8,000 Euro
- Workpackage Leader for Embedded Systems for Autonomous Robotics within the METU-ISTEC project (an Specific Support Action type project funded by TUBITAK). Workpackage budget: 15,000 Euro.
- Third-party for the SEE-GRID2 project through TUBITAK ULAKBIM. Within this project we developed an application for parallel execution of artificial evolution
