Robotics R&D · Research & industry
Perception, motion planning and control for legged and field robots, from simulation to deployment.
Six lines of work, 2019–2026, in research labs and industry projects: from simulation studies to systems tested on real hardware.
Controllers are trained and tested in Gazebo and MuJoCo: deep-RL trotting on slopes up to 15°; DDPG, PPO and SAC compared; a Cost-of-Transport reward. Even camera placement for gait capture is chosen in simulation.
Code meets hardware: an EtherCAT master and ROS/ROS 2 middleware for the Laelaps II quadruped, a low-cost gantry-harvester prototype with 360° RGB-D scanning and visual servoing, a vineyard robot with three stereo cameras.
On test beds and in the field: a synthetic vineyard, 3D-printed mushrooms on real soil, die-cast parts in a two-cobot cell, e-waste on a moving conveyor; deployments in agriculture, energy, healthcare logistics and industry.
The same problems, revisited: slope climbing (2015), compliant terrain (2018), deep-RL slope handling (2020, CLAWAR prize), energy-efficient trotting (2022–2025); gait analysis from OpenPose to multi-camera meshes to a minimal model (2019–2025).
Real-time replanning without GPS from a local map that grows with vehicle speed; collision-free coverage of a disaster area by a UAV swarm; autonomous inspection of extra-high-voltage substations; and visual-inertial odometry that keeps tracking inside pipes by spraying its own fluorescent landmarks.
Setupspeed-adaptive local ESDF map · UAV swarm formations · ENORASI substation robot · in-pipe VIO with optimal IMU placement
Finding battery-containing items in a stream of e-waste so a robot can pick them out, and catching surface and thread defects on aluminum die-cast automotive parts in real time.
Setupdual-energy X-ray · YOLO + U-Net · Delta robot over a moving conveyor · two-cobot inspection cell
Autonomous navigation, high-fidelity 3D reconstruction and canopy growth assessment for vineyard inspection; multi-camera visual SLAM that stays robust when one camera’s view is largely blocked.
Setupsynthetic vineyard test bed · three stereo cameras · AprilTag loop closure
Best Agri-Robotics Paper Finalist · IROS 2024A harvesting robot that fits the shelves mushroom farms already have: it finds each mushroom, estimates its pose and steers the gripper onto it at the right angle, accurately and at low cost. Part of the EU-funded SoftGrip project.
Setuplow-cost gantry on existing farm shelves · 3D-printed mushrooms on real soil · 360° RGB-D scan · YOLOv5 + model-fit pose · visual servoing
Best Paper Finalist · AIM 2023Trotting controllers for the Laelaps II quadruped that combine deep reinforcement learning with toe trajectory planning to handle slopes. A companion study compared three learning algorithms; later work rewarded energy-efficient trotting.
SetupLaelaps II quadruped model in Gazebo, slopes up to 15° · MuJoCo for Cost-of-Transport runs · DDPG vs PPO vs SAC
Second Prize · CLAWAR 2020Low-cost alternatives to marker-based motion capture: measuring limbs from 2D pose estimates, reconstructing 3D human meshes from several ordinary cameras, and estimating joint angles, torques and ground reaction forces from a single stereo camera with a minimal human model.
SetupOpenPose · multi-camera rig with simulation-guided placement · stereo camera · minimal 2D human model
Perception and motion planning for legged robots.
Projects ARGOS (quadruped for precision agriculture), INTELLICONT, DAAD, THALES and HEXATERRA.
Led technical teams and mentored junior engineers, specializing in robotics software, AI/ML integration and simulation environments.
Full project lifecycle delivery of robotics and AI-driven solutions, from soft harvesting grippers and multi-drone energy inspection to X-ray battery detection and pipeline inspection robots.
ENDORSE: indoor robotic fleets for logistics in healthcare and commercial spaces.
Athanasios (Thanasis) Mastrogeorgiou works at the boundary of robotics research and industry. He is a Senior Robotics Researcher at HERON and a Ph.D. candidate in Mechatronics; from 2018 to 2025 he worked in parallel in industry, as Robotics Software Engineer at IKNOWHOW and CiTARD, Senior Robotics Engineer at RobotiCo and Delivery Manager at Tech Hive Labs, where he managed nine robotics and AI projects end to end.
He takes perception, motion planning and control from high-fidelity simulation to field deployments in agriculture, healthcare logistics, energy infrastructure and industrial automation; on the way, the work passes through test beds such as a synthetic vineyard, 3D-printed mushrooms on real soil and Gazebo slopes of up to 15°. Since 2014 he has worked with the Legged Robots Team on the Laelaps II quadruped, from its EtherCAT master to deep-RL trotting, and on vineyard inspection with multi-camera SLAM; within the EU-funded SoftGrip project he helped build a low-cost mushroom-harvesting gantry.
National Technical University of Athens · 2016 – Present · Supervisor: Prof. E. G. Papadopoulos
National Technical University of Athens · 2011 – 2013
University of Edinburgh · 2008 – 2009
University of Macedonia · 2003 – 2008
Gazebo · MuJoCo · Webots · MSC Adams
ROS/ROS 2 · C++ · Deep reinforcement learning · Visual SLAM
Embedded systems · EtherCAT · Stereo and RGB-D cameras · IMUs
Greek (native) · English (C2) · German (B2)
Martial arts · Photography · Mountain climbing · FPV drone builds
Taught computer essentials at ARSIS (2017); Hellenic Post Office of Hydra, “Holidays and volunteering” program (2011).