Robot Learning Researcher mehmetturanyardimci@hotmail.com | LinkedIn
I'm a Computer Engineering graduate from Çukurova University (2025) with a strong focus on robotics, reinforcement learning, and autonomous navigation. I specialize in implementing state-of-the-art RL algorithms from scratch and deploying them on robotic platforms.
Currently working on:
- Quadruped Locomotion with Isaac Lab
- PPO implementations for continuous control
- Local Path Planning benchmark (BARN dataset)
Robotics & Simulation: Isaac Sim • MuJoCo • Gazebo • ROS/ROS2
Machine Learning: PyTorch • NumPy • Deep RL (PPO, SAC)
Computer Vision: YOLOv4-8 • OpenCV
Platforms: Jetson Nano • Pixhawk • Linux
From-scratch implementation of PPO algorithm for the MuJoCo Ant-v5 environment.
Key Features:
- Pure NumPy & PyTorch (no stable-baselines3)
- Parallel environments (16 envs)
- TensorBoard integration
- Achieved 2700+ reward in 8M steps
Technical Highlights:
- GAE (λ=0.95), reward normalization
- Clipped surrogate objective
- Learning rate annealing
- Exploration decay monitoring
Interactive Streamlit app demonstrating real-time RL training on CartPole.
Features:
- Watch the agent learn live in your browser
- Adjustable hyperparameters
- Educational tool for understanding Actor-Critic methods
Comprehensive benchmark of ROS navigation planners using the BARN dataset.
Planners Tested: TEB, DWA, MPC, Lattice
Status: Under publication review
Focus: Real-world applicability and computational efficiency
UAV Team Lead | 1.5 Adana AGM ALKAR (3 years)
- Led autonomous drone project with YOLO object detection
- Integrated ROS1/2, Gazebo, Pixhawk flight controller
- Deployed YOLOv7 on Jetson Nano for real-time inference
Currently diving deep into:
- Isaac Sim for quadruped locomotion (Anymal, Unitree A1, Spot)
- Domain Randomization for sim-to-real transfer
- Hierarchical RL for complex robotic behaviors
I'm open to research collaborations and R&D opportunities in robotics and AI.
Feel free to reach out: mehmetturanyardimci@hotmail.com