Emad Razavi

Emad Razavi

MSc in Robotics Engineering from the University of Genoa.

About

I am a robotics engineer with an MSc in Robotics Engineering from the University of Genoa, and thesis work at the Dynamic Legged Systems lab at the Istituto Italiano di Tecnologia.

My background spans real robot autonomy and low level robotics engineering. At IIT DLS I built and deployed an end to end ROS 2 stack on a Boston Dynamics Spot, Nav2 for indoor navigation, odometry from 2D LiDAR plus a tracking camera, a confirmed only semantic map from RGB D detections for object goal navigation, and voice commanded goals using Whisper and CLIP embeddings. And I also come from embedded systems and electronics, including drone flight controllers and wireless links, STM32 based robots with IMU fusion, and RoboCup holonomic robots with custom PCBs and firmware.

Research and Thesis

My MSc thesis is a semantic object-goal navigation pipeline for the Boston Dynamics Spot quadruped, built in ROS 2.

Stage 1: Pre-run mapping

Build a 2D occupancy map and collect object candidates from RGB-D detections. Only confirmed objects are kept, so the semantic layer stays stable and usable for goal selection. The semantic mapping component is described in this paper: Semantic Object Mapping on a Quadruped Robot.

Stage 2: Object-goal navigation

Navigate to selected objects using Nav2, with a voice interface that maps spoken commands to targets using Whisper and CLIP embeddings. The navigation pipeline is documented in my thesis. thesis.

What I have worked on

  • End-to-end autonomy on Spot: drivers, sensor setup, data flow, mapping, localization, Nav2 tuning, and on-robot deployment.
  • Semantic mapping: RGB-D detections fused into a confirmed-only object database in the map frame for goal selection.
  • System engineering: Docker setups, launch and test scripts, rosbags, and debugging TF, timing, QoS, and calibration issues.
  • Earlier work: embedded and control systems with STM32, PCB design, and holonomic robot platforms.

Technical focus

Robotics

ROS 2, Nav2, TF, SLAM, mapping and localization, planning and control, field debugging

Perception

RGB-D pipelines, object detection, sensor fusion, OpenCV, point cloud workflows

Engineering

C++ and Python, Git, Linux, Docker, Gazebo and MuJoCo, reproducible runs and logging

Education

MSc Robotics Engineering, University of Genoa (2023–2025)

Thesis: Semantic Object-Goal Navigation on a Quadruped Robot in Known Environments (IIT DLS)

BSc Electrical Engineering, Qazvin Azad University (2018–2023)

Thesis: Optimal PID Controller Tuning for Time-Delayed Systems Using Ant Colony Optimization

Bachelor Thesis: ACO-Based PID Tuning

My bachelor thesis developed a metaheuristic optimization approach for tuning PID controllers in time-delayed systems. Using Ant Colony Optimization (ACO), the method minimizes the Integral Time Absolute Error (ITAE) to achieve fast settling times and reduced overshoot.

The algorithm uses pheromone-based exploration to navigate the discrete PID parameter space (Kp, Ki, Kd), with roulette-wheel selection and adaptive pheromone decay. Results demonstrated significant performance improvement over the untuned baseline, with faster convergence and better handling of transport delays.

View on GitHub