Basheer Al-Tawil

Basheer Al-Tawil

PhD researcher in robotics at Otto von Guericke University Magdeburg

Robots that map, understand and move among people.

I build the perception and navigation software that lets a mobile robot keep an accurate map while people walk through it, recognise what they are doing, and plan a path that respects their space.

See the research Download CV
Click the map to send the robot somewhere. robot person social cost planned path

From sensor data to polite motion

My work follows the data through a mobile robot, in five steps.

  1. 1SenseMulti-sensor fusion for 2D mapping
  2. 2MapVisual SLAM in dynamic scenes
  3. 3UnderstandHuman action, trajectory and engagement
  4. 4NavigateSocially-aware navigation
  5. 5ReasonLLMs for robot task control

Results so far

How these were achieved
  • 96 %

    lower trajectory error than ORB-SLAM3 in scenes with moving people, from 0.404 m to 0.015 m on TUM RGB-D.

    SAR-SLAM, Robotics 2026
  • 96.6 %

    action-recognition accuracy on UCF-101 with a ResNet-18, Bi-LSTM and attention model that runs on a mobile robot.

    Sensors 2025
  • 95 %

    of navigation goals reached after fusing four RGB-D cameras with two LiDARs for 2D mapping.

    Sensors 2025
  • 140+

    citations for my review of visual SLAM methods for robotics, my most-cited paper.

    Frontiers in Robotics and AI 2024

Selected publications

All 13 publications

Open-source projects

All 8 projects
Since 2019

RoboCraft: a reconfigurable three-arm SCARA cell

Three SCARA robots that work alone or lock onto a hexagonal platform and become one parallel robot.

  • ROS 2 Humble
  • Gazebo Fortress
  • ros2_control
  • C++17
Since 2020

AgroBot: a robot arm for greenhouse work

A 4-axis arm on a rail trolley that harvests strawberries, transplants seedlings, inspects plants and pulls weeds.

  • ROS 2 Humble
  • Gazebo Fortress
  • ros2_control
  • C++17
Camera view from a mobile robot showing a tracked person with a skeleton overlay, the observed walking path in blue and the predicted path in green
2026

TrajMamba: predicting where people will walk

A 262 K-parameter Mamba model that predicts the next three seconds of a person’s path from an RGB-D camera, in real time on a Jetson Orin.

  • ROS 2 Humble and Jazzy
  • PyTorch
  • Mamba
  • MediaPipe Pose

Tutorials and notes

All 13 posts

  • Introduction to robotics and AI

    What a robot is, how AI differs from machine learning, where you already meet both in daily life, and the parts every robot is built from. Day 1 of the AIBOMECH Ramadan series.

    AI5 min read

  • Visual SLAM from theory to implementation

    A 13-chapter study guide to Visual SLAM: pose representation, Lie groups, camera models, nonlinear optimisation, feature-based and direct odometry, bundle adjustment, loop closure and dense mapping, with C++ sketches.

    SLAM19 min read

  • Keypoints, descriptors and feature matching for robotics

    How Harris, FAST, SIFT and ORB find and describe image features, how nearest-neighbour matching and the ratio test pair them, and how a SLAM front end turns matches into robot motion. With Python and OpenCV code.

    Computer vision21 min read

Work with me

I am open to research collaborations, industry R&D roles, postdoc positions and invited talks on mobile robotics, SLAM, perception and human-robot interaction.

basheer.al-tawil@ovgu.de Send a message

Browse by topic