Physical AI · Humanoid Robotics · VLA Models · ROS 2 · MPC · Sensor Fusion · Autonomous Navigation
I work on robotics and autonomous systems, with a current focus on Physical AI for industrial humanoid robots.
I am currently one of the co-founders of Talos Robotics AI, where we are building TalOS, a Physical AI operating system that helps humanoid robots understand tasks, reason about the environment, validate actions in simulation, and execute useful work in real industrial settings.
My background combines optimal control, state estimation, robot navigation, ROS 2 systems, and machine learning, with completed work on UAV autonomy and current research/development on Vision-Language-Action systems for humanoid robots.
At Talos Robotics AI, I work on enhancing humanoid Vision-Language-Action systems along with Reinforcement Learning for industrial applications.
My current interests include:
-
Physical AI for industrial robotics
- Turning high-level human instructions into executable robot behaviors
- Bridging perception, reasoning, simulation, and real-world deployment
-
Humanoid robotics
- Task execution on real industrial floors
- Multi-modal robot understanding for tools, objects, people, and workspaces
-
Vision-Language-Action models
- Grounding language commands into robot actions
- Improving robustness, safety, and task generalization
- Enhancing humanoid autonomy for practical industrial workflows
-
Sim-to-real robotics
- Validating behaviors in simulation before deployment
- Building reliable pipelines from task description to real robot execution
Co-founder — Talos Robotics AI
Talos Robotics AI is building a Physical AI OS for humanoid robots, designed to make robots useful in real-world industrial environments.
The goal is to let users describe a task in plain language, test it in simulation, and deploy it on a physical humanoid robot with perception, planning, and safety integrated into one platform.
Accepted at IEEE ICRA 2026 Xplore Workshop — Oral Presentation
This work presents a robot-agnostic autonomous navigation framework that combines:
-
LiDAR-based Gaussian occupancy representation
-
A* rolling-horizon planning
-
Nonlinear Model Predictive Control
-
Bayesian Optimization for MPC parameter tuning
-
Simulation-to-real deployment on a Unitree Go2 quadruped robot
-
Physical AI & humanoid robotics
- Vision-Language-Action systems for industrial humanoids
- Task grounding, robot reasoning, and real-world deployment
-
Autonomous navigation
- MPC-based planning and control
- Bayesian Optimization for controller tuning
- Sim-to-real validation on legged robots
-
UAV trajectory planning & control
- Completed work on MPC-based quadrotor trajectory generation
- Constraint-aware planning for real-world drone dynamics
-
Sensor fusion & state estimation for UAV's
- Extended Kalman Filters
- Multi-sensor fusion with IMU, GPS, magnetometer, and barometer data
-
ROS 2 & simulation
- ROS 2 Humble development
- Gazebo-based testing, simulation, and integration
-
Visual SLAM & perception
- Stereo / RGB-D visual odometry
- Mapping and pose-estimation pipelines
Bayesian Optimization for Learning Nonlinear MPC in Autonomous Agent Navigation
Robot-agnostic autonomous navigation framework using rolling-horizon planning, nonlinear MPC, Bayesian Optimization, and sim-to-real validation on a Unitree Go2 quadruped robot.
➡️ https://github.com/talos-robotics-ai/Go2_navigation
Completed UAV trajectory planning project using MPC-based approaches for quadrotor navigation with ROS 2, Gazebo, and PX4.
➡️ https://github.com/Relo02/Drone-optimal-trajectory
Completed EKF-based sensor fusion framework for quadrotor state estimation using multiple onboard sensors.
➡️ https://github.com/Relo02/Quadcopter-Sensor-Fusion
Completed ROS 2 Humble simulation environment for UAV control-system experimentation, obstacle awareness, and smart landing procedures.
➡️ https://github.com/FALCOdrone/Ros-2-Environment
Academic and practical work on neural networks and deep learning concepts.
➡️ https://github.com/Relo02/Artificial-neural-networks-and-deep-learning-
Visual SLAM and VIO pipelines using stereo / RGB-D sensors for pose estimation and mapping.
➡️ https://github.com/palingenesys/Visual-Slam
- Robotics: ROS 2 Humble, Gazebo, PX4, autonomous navigation
- Physical AI: humanoid robotics, VLA systems, Reinforcement Learning, sim-to-real deployment
- Control & Planning: MPC, nonlinear MPC, optimal control, trajectory optimization
- Optimization: Bayesian Optimization, Gaussian Processes, TPE-based tuning
- Estimation & Perception: EKF, sensor fusion, Visual SLAM, VIO
- Sensors: IMU, GPS, LiDAR, stereo cameras, RGB-D cameras
- ML/DL: PyTorch, neural networks, deep learning
- Development: Python, C++, MATLAB, Docker, Linux, Arduino
- Email: ortolore@gmail.com
- LinkedIn: https://www.linkedin.com/in/lorenzo-ortolani-6135b7240/
- Talos Robotics AI: https://talosrobotics.ai/
⭐ I enjoy turning robotic systems, Physical AI, and complex autonomous software into real-world robotic capabilities.
