Research

Featured Projects

A Safety-Aware Shared Autonomy Framework with BarrierIK Using Control Barrier Functions

BarrierIK system overview Shared autonomy blends operator intent with autonomous assistance, but in cluttered environments, linearly blending individually collision-free commands can still produce unsafe motion. Most prior approaches only enforce obstacle avoidance as a soft constraint through potentials or cost terms. In this work, we introduce BarrierIK, which embeds control barrier functions (CBFs) directly into the inverse kinematics (IK) layer of shared autonomy, enforcing safety as a hard constraint while remaining compatible with standard IK objectives such as pose tracking, joint limits, and smoothness. We evaluate BarrierIK in simulation on cluttered manipulation scenes and in a VR teleoperation user study comparing pure teleoperation with shared autonomy. Across conditions, filtering blended commands through CBFs at the IK layer reduces safety-violation time and increases minimum clearance while preserving task performance; participants also reported higher perceived safety and trust, lower interference, and an overall preference for shared autonomy with our safety filter.

    My Contributions:
  • Formulated and implemented BarrierIK, a control-barrier-function-constrained inverse kinematics solver that enforces post-blend safety as a hard inequality constraint alongside standard IK objectives.
  • Designed and ran the simulation benchmarks in cluttered environments, evaluating safety-violation time and minimum clearance against baseline shared-autonomy blending.
  • Designed and conducted the VR teleoperation user study comparing pure teleoperation with CBF-filtered shared autonomy.
  • Analyzed the objective (clearance, violation time, task performance) and subjective (trust, perceived safety, interference) study results.

B. Guler, K. Pompetzki, Y. Sun, S. Manschitz, and J. Peters, “A Safety-Aware Shared Autonomy Framework with BarrierIK Using Control Barrier Functions,” IEEE International Conference on Robotics and Automation (ICRA), 2026. Available:
arXiv Project Page

AssistDLO: Assistive Teleoperation for Deformable Linear Object Manipulation

Manipulating Deformable Linear Objects (DLOs) such as ropes and cables is challenging in robotics due to their infinite-dimensional configuration space and complex nonlinear dynamics, and in teleoperation, depth uncertainty further hinders state perception and reaction. AssistDLO addresses this as an assistive teleoperation framework that combines real-time multi-view state estimation, visual assistance (VA), and a geometry-aware shared-autonomy controller based on Control Barrier Functions (SA-CBF). Unlike traditional shared-autonomy methods that rely on simple geometric attractors and can fail to preserve DLO geometry, SA-CBF acts as a geometry-aware funnel that facilitates precise grasping while preserving the operator's high-level authority. The framework was evaluated in a bimanual knot-untangling user study (N = 22) using ropes of varying length and rigidity: SA-CBF provides the strongest gains for novice users, acting as a skill equalizer that increases task success from 71% to 88%, and is most effective on stiffer ropes, while expert users and highly compliant, long ropes benefit more from visual assistance — showing that effective DLO teleoperation requires adaptive, user- and material-aware shared autonomy rather than a fixed assistance strategy.

    My Contributions:
  • Designed the AssistDLO framework, combining real-time multi-view DLO state estimation with a geometry-aware shared-autonomy controller.
  • Formulated the SA-CBF shared-autonomy controller as a geometry-aware safety funnel around the estimated DLO state for precise, collision-aware grasping.
  • Designed and ran the bimanual knot-untangling user study (N = 22) across ropes of varying length and rigidity, comparing visual assistance and shared autonomy against unassisted teleoperation.
  • Analyzed how assistance effectiveness depends on operator expertise and DLO material properties.

B. Guler, S. Manschitz, K. Pompetzki, and J. Peters, “AssistDLO: Assistive Teleoperation for Deformable Linear Object Manipulation,” arXiv preprint arXiv:2605.06323, 2026. Available:
arXiv

Robot-Assisted Drilling on Curved Surfaces with Haptic Guidance under Adaptive Admittance Control

Drilling a hole on a curved surface with a desired angle is prone to failure when done manually, due to the difficulties in drill alignment and also inherent instabilities of the task, potentially causing injury and fatigue to the workers. On the other hand, it can be impractical to fully automate such a task in real manufacturing environments because the parts arriving at an assembly line can have various complex shapes where drill point locations are not easily accessible, making automated path planning difficult. In this work, an adaptive admittance controller with 6 degrees of freedom is developed and deployed on a KUKA LBR iiwa 7 cobot such that the operator is able to manipulate a drill mounted on the robot with one hand comfortably and open holes on a curved surface with haptic guidance of the cobot and visual guidance provided through an AR interface.

    My Contributions:
  • Developed an admittance controller in C++ using Fast Robot Interface (FRI) of KUKA iiwa co-bot.
  • Developed an mixed reality (MR) interface in C#, working with Microsoft HoloLens, using Unity Game Engine and Microsoft's Mixed Reality Toolkit (MRTK).
  • Implemented a Kalman Filter to track 6-DoF movements of HoloLens and the workpiece being drilled via visual fiducial markers (AprilTag) attached to them.
  • Constructed a 3D mesh model of the workpiece using a depth camera (Microsoft Kinect).
  • Established a TCP communication between the robot and HoloLens.
A. Madani, P. P. Niaz, B. Guler, Y. Aydin, and C. Basdogan, “Robot-Assisted Drilling on Curved Surfaces with Haptic Guidance under Adaptive Admittance Control” IROS, 2022. Available:

An adaptive admittance controller for collaborative drilling with a robot based on subtask classification via deep learning

In this paper, we propose a supervised learning approach based on an Artificial Neural Network (ANN) model for real-time classification of subtasks in a physical human-robot interaction (pHRI) task involving contact with a stiff environment. In this regard, we consider three subtasks for a given pHRI task: Idle, Driving, and Contact. Based on this classification, the parameters of an admittance controller that regulates the interaction between human and robot are adjusted adaptively in real-time to make the robot more transparent to the operator (i.e., less resistant) during the Driving phase and more stable during the Contact phase. The Idle phase is primarily used to detect the initiation of the task.

    My Contributions:
  • Developed a control architecture for regulating the physical interactions between human and a robot (pHRI) via an admittance controller implemented on KUKA iiwa co-bot.
  • Performed a linear time-varying stability analysis for the pHRI system.
  • Developed a deep learning (DL) model to classify the subtasks of a task using the Tensorflow framework in Python.
  • Developed methods to adapt the parameters of the admittance controller based on the outcome of the DL model.
  • Developed a graphical user interface (GUI) in C# for reading and recording data from ATI force sensors.
  • Designed pHRI experiments, collected data, and performed statistical analysis on the data.
B. Guler, P. P. Niaz, A. Madani, Y. Aydin, and C. Basdogan, “An adaptive admittance controller for collaborative drilling with a robot based on subtask classification via deep learning” Mechatronics, vol. 86, p. 102851, 2022. Available:

Autonomous Manipulation and SLAM with the Hybrid Mobile Manipulator

This study demonstrates the soft robotic as a part of the mobile manipulator that executes the defined tasks, which are listed below:

  • Autonomous Navigation using SLAM.
  • The visual recognition of the marked boxed with ARtags.
  • Pick&Place manipulation of the objects located on top of the marked boxes.
    My Contributions:
  • Designed a serial robotic manipulator using Siemens NX and manufactured it using 3-D printing technology.
  • Designed a soft gripper using PneuNets, and analyzed by utilizing finite element methods using Abaqus.
  • Integrated the robotic manipulator with a mobile platform.
  • Developed the ROS packages for the manipulator in Python.
  • Tracked the positions of the marked boxes using visual fiducial markers (ARTag) attached to the box using OpenCV.
  • Implemented a SLAM algorithm from ROS Navigation Stack to estimate the pose of the mobile platform using LIDAR and IMU.

Autonomous Visual Fiducial Markers Recognition and Tracking

Approaching the Visual Fiducial Marker

Inverse Kinematics

Deep Reinforcement Learning to Optimize Task Performance in Human-Robot Co-Manipulation

In industry, complex tasks encourage human operators to work together with robots, and carrying heavy objects is one of them. Using the human's dexterity and intelligence and the robot's precision, we can drive and park heavy objects where the human operator wants without the programming in advance. This study's initial aim is to increase efficiency by understanding human motion intention. However, the human force's response is observed by delay due to the object's weight.Therefore, estimated human intention can help make decisions on amplifying or attenuating the human force. Constantly scaling in one way is not efficient for human effort because of the different phases of the co-manipulation task. Amplification is preferable during driving when attenuation becomes more of an issue for parking. The deep learning model provides an adaptive force scaling factor by referring to human motion intention without guaranteeing optimal human effort and comfort. The model-free optimal control algorithm, deep reinforcement learning, adjusts the existing adaptive force scaling factor to decrease human effort and increase comfort.

    My Contributions:
  • Developed an experimental set-up for human-robot co-manipulation of a heavy object.
  • Developed a deep learning (DL) model using Tensorflow in Python to estimate human intention to accelerate/decelerate the manipulated object.
  • Developed a deep reinforcement learning (DRL) model, working with the DL model, to optimize task performance based on minimum jerk and human effort.
  • Designed pHRI experiments, collected data, and performed statistical analysis on the data.

B. Guler, “Deep Reinforcement Learning to Optimize Task Performance in Human-Robot Co-Manipulation,” M.S. Thesis, Koç University, 2023.

Full Publication List

An up-to-date list is also maintained on my Google Scholar profile.

Journal Articles

  1. B. Guler, P. P. Niaz, A. Madani, Y. Aydin, and C. Basdogan, “An adaptive admittance controller for collaborative drilling with a robot based on subtask classification via deep learning,” Mechatronics, vol. 86, p. 102851, 2022.
    arXiv

Conference Papers

  1. B. Guler, K. Pompetzki, Y. Sun, S. Manschitz, and J. Peters, “A Safety-Aware Shared Autonomy Framework with BarrierIK Using Control Barrier Functions,” in IEEE International Conference on Robotics and Automation (ICRA), 2026.
    arXiv Project Page
  2. S. Manschitz, B. Guler, W. Ma, and D. Ruiken, “Sampling-Based Grasp and Collision Prediction for Assisted Teleoperation,” in IEEE International Conference on Robotics and Automation (ICRA), 2025.
    arXiv
  3. B. Guler, K. Pompetzki, S. Manschitz, and J. Peters, “Towards Assistive Teleoperation for Knot Untangling,” in German Robotics Conference (GRC), 2025.
    PDF
  4. A. Madani, P. P. Niaz, B. Guler, Y. Aydin, and C. Basdogan, “Robot-Assisted Drilling on Curved Surfaces with Haptic Guidance under Adaptive Admittance Control,” in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022.
    arXiv

Workshop Papers

  1. G. Wigginghaus, T. Missal, B. Guler, S. Manschitz, and J. Peters, “Learning Sim-Grounded Policies for Bimanual Rope Manipulation from Human Teleoperation Data,” in 1st ICRA Workshop on Beyond Teleoperation, 2026.
    arXiv

Preprints

  1. B. Guler, S. Manschitz, K. Pompetzki, and J. Peters, “AssistDLO: Assistive Teleoperation for Deformable Linear Object Manipulation,” arXiv:2605.06323, 2026.
    arXiv
  2. T. Missal, L. Domingues, B. Guler, S. Manschitz, J. Peters, and P. D. P. Costa, “RopeDreamer: A Kinematic Recurrent State Space Model for Dynamics of Flexible Deformable Linear Objects,” arXiv:2604.28161, 2026.
    arXiv

Thesis

  1. B. Guler, “Deep Reinforcement Learning to Optimize Task Performance in Human-Robot Co-Manipulation,” M.S. Thesis, Koç University, 2023.