Eren Erdal Aksoy

About

Eren Erdal Aksoy is a prominent robotics and computer vision researcher whose work sits at the intersection of machine learning, semantic reasoning, and robot manipulation. He is best known for pioneering the concept of **semantic object-action relations (SOARs)** and semantic scene graphs, frameworks that enable robots to observe, understand, and reproduce human manipulation behaviors with remarkable generality. Aksoy's most influential contribution, "Learning the Semantics of Object–Action Relations by Observation" (2011, 180 citations), introduced a novel relational representation that captures the critical interactions between objects during manipulation, forming the foundation for much of his subsequent research. Building on this, he developed ontologies and library-based frameworks for categorizing and executing manipulation actions, culminating in practical robot execution systems capable of learning by demonstration across diverse tasks. His research extends into part-based grasp planning and cognitive architectures for assembly tasks, demonstrating the real-world applicability of his theoretical contributions. A notable departure into legged robotics — visual terrain classification for energy-efficient hexapod locomotion (51 citations) — highlights his broader technical versatility. With over 670 cumulative citations across his top works, Aksoy has made a lasting impact on how robots perceive, learn, and generalize manipulation skills, offering valuable tools for the next generation of intelligent, task-adaptive robotic systems.

Research Focus

Key Achievements

Top Papers

1. 1
   Learning the semantics of object–action relations by observation

   180 citations · 2011
2. 2
   A Simple Ontology of Manipulation Actions Based on Hand-Object Relations

   87 citations · 2013
3. 3
   Categorizing object-action relations from semantic scene graphs

   80 citations · 2010
4. 4
   Teaching a Robot the Semantics of Assembly Tasks

   64 citations · 2017
5. 5
   Part-based grasp planning for familiar objects

   56 citations · 2016
6. 6
   Visual terrain classification for selecting energy efficient gaits of a hexapod robot

   51 citations · 2013
7. 7
   Action sequence reproduction based on automatic segmentation and Object-Action Complexes

   46 citations · 2013
8. 8
   Model-free incremental learning of the semantics of manipulation actions

   38 citations · 2014
9. 9
   Library of actions: Implementing a generic robot execution framework by using manipulation action semantics

   36 citations · 2019
10. 10
    Toward a library of manipulation actions based on semantic object-action relations

    35 citations · 2013

Key Collaborators

Contact & Links