Publications

Papers on SLAM, perception, understanding people and robot arms. Each entry links to the paper, and to the code where it is public.

11published works 201citations Counts from Google Scholar, October 2026.

2026

  • SAR-SLAM: Semantic-Aware Recognition for Dynamic SLAM in Robotic Applications

    Basheer Al-Tawil, Magnus Jung, Thorsten Hempel, Ayoub Al-Hamadi

    Robotics 15(7), 136, 2026

    An RGB-D SLAM framework for scenes with moving people. YOLOv8 segmentation proposes dynamic objects, a RANSAC homography verifies which ones really move, and an adaptive fusion step filters their features before ORB-SLAM3. Absolute trajectory error drops by up to 96 % on dynamic TUM RGB-D sequences.

    SAR-SLAM pipeline. An RGB-D frame goes through semantic segmentation and motion direction estimation, the two are fused, and ORB-SLAM3 tracking produces the trajectory and the 3D map.
  • Deep Learning-Based Gaze Estimation: A Review

    Ahmed A. Abdelrahman, Basheer Al-Tawil, Ayoub Al-Hamadi

    Robotics 15(4), 69, 2026

    A review of deep-learning gaze estimation, from conventional methods to appearance-based networks, comparing input features, architectures and preprocessing and naming the open gaps.

    DOI: 10.3390/robotics15040069SJR Q1Cited by 3
    Gaze estimation geometry. Left, the camera, screen and head coordinate frames with the gaze direction. Right, the eyeball model with the gaze vector reaching the point of gaze.
  • REMA-Net: Real-Time Human Action Recognition via Caption-Guided Dual-Stream Attention

    Basheer Al-Tawil, Magnus Jung, Ayoub Al-Hamadi

    SSRN preprint

    A lightweight dual-stream network that pairs an RGB stream with a BLIP caption stream through cross-modal knowledge distillation. 96.8 % on UCF-101, 73 % on HMDB-51 and 34.8 % on HAA-500, with 11.82 ms latency on a Jetson Nano.

    Preprint, under reviewRead the preprint
    REMA-Net overview. The training pipeline with a ResNet-18 student and a caption-based teacher, two baseline architectures, and the single-stream pipeline deployed on the robot.

2025

  • Toward Truly Intelligent Autonomous Systems: A Taxonomy of LLM Integration for Everyday Automation

    Magnus Jung, Thorsten Hempel, Basheer Al-Tawil, Qiaoyue Yang, Sven Wachsmuth, Ayoub Al-Hamadi

    Robotics, Computer Vision and Intelligent Systems (ROBOVIS 2025), pp. 74–91, 2025

    Three levels of wiring a large language model into a robot: open-loop, closed-loop and fully LLM-driven. Tested on the TIAGo robot and in simulation with GPT-4 and Vicuna.

  • Eye Contact Based Engagement Prediction for Efficient Human–Robot Interaction

    Magnus Jung, Ahmed Abdelrahman, Thorsten Hempel, Basheer Al-Tawil, Qiaoyue Yang, Sven Wachsmuth, Ayoub Al-Hamadi

    Complex & Intelligent Systems 11, 286, 2025

    Predicts whether a person is engaged with a robot from just two features, eye contact and distance, instead of more than a hundred. Real-time, with 80.73 % accuracy on the UE-HRI benchmark.

    DOI: 10.1007/s40747-025-01902-zSJR Q1Cited by 10
  • Multi-Head Attention-Based Framework with Residual Network for Human Action Recognition

    Basheer Al-Tawil, Magnus Jung, Thorsten Hempel, Ayoub Al-Hamadi

    Sensors 25(9), 2930, 2025

    ResNet-18 features, a Bi-LSTM and multi-head attention, with optical-flow frame selection, reach 96.60 % accuracy on UCF-101. Validated on a mobile robot for real-time use.

    DOI: 10.3390/s25092930CodeSJR Q1Cited by 10
    Figure from the paper: Multi-Head Attention-Based Framework with Residual Network for Human Action Recognition
  • Mobile Robot Navigation with Enhanced 2D Mapping and Multi-Sensor Fusion

    Basheer Al-Tawil, Adem Candemir, Magnus Jung, Ayoub Al-Hamadi

    Sensors 25(8), 2408, 2025

    RGB-D point clouds are projected to 2D, denoised and fused with LiDAR scans. Enhanced Gmapping (EGM) adds adaptive resampling and degeneracy handling, which makes mapping and navigation more robust indoors.

    DOI: 10.3390/s25082408SJR Q1Cited by 20
    Figure from the paper: Mobile Robot Navigation with Enhanced 2D Mapping and Multi-Sensor Fusion
  • Joint Estimation of Affective and Attentional States from Egocentric Video Input for Human-Robot Interaction

    Magnus Jung, Basheer Al-Tawil, Sven Wachsmuth, Ayoub Al-Hamadi

    Submitted to a Springer Nature journal

    Under review

2024

  • A Review of Visual SLAM for Robotics: Evolution, Properties, and Future Applications

    Basheer Al-Tawil, Thorsten Hempel, Ahmed Abdelrahman, Ayoub Al-Hamadi

    Frontiers in Robotics and AI 11, 1347985, 2024

    Traces how visual SLAM evolved, gives criteria for choosing a method for a given robot, and compares the major systems by their ROS support and benchmark results.

    DOI: 10.3389/frobt.2024.1347985SJR Q2Cited by 140
    Figure from the paper: A Review of Visual SLAM for Robotics: Evolution, Properties, and Future Applications

2023

  • Design and Analysis of a Four DoF Robotic Arm with Two Grippers Used in Agricultural Operations

    Basheer Altawil, Fatih Cemal Can

    International Journal of Applied Mathematics Electronics and Computers 11(2), pp. 79–87, 2023

    A 4-DoF arm with two grippers for greenhouse tasks. Kinematics by the Denavit–Hartenberg method, dynamics by Lagrangian mechanics, and integration with ROS.

    Figure from the paper: Design and Analysis of a Four DoF Robotic Arm with Two Grippers Used in Agricultural Operations
  • Automating Farming Operations Using Robotic Technologies

    Basheer Altawil

    M.Sc. thesis, İzmir Kâtip Çelebi University, 2023

    A survey of agricultural robots for monitoring, spraying, harvesting and transport, and a ROS ecosystem that ties middleware, hardware and edge IoT devices together for grasping tasks in farming.

2022

  • Automating Farming Operations Using Robotics Technologies

    Basheer Altawil et al.

    ICAT conference, pp. 353–359, 2022

2019

  • Development of a Multi Degrees of Freedom Redundant-Parallel Manipulator

    Basheer Altawil et al.

    ISSC conference, pp. 540–547, 2019

    A redundant planar parallel manipulator with six motors and three degrees of freedom whose serial arms can detach and work alone.

    Code

Browse by topic