Robot Learning :

• Analysis of Sim-to-Real strategies with Domain Randomization techniques - group project : analysis of sim-to-real transfer in robotics by addressing the reality gap using Domain Randomization. Comparison of Uniform Domain Randomization (UDR), which improves robustness via fixed randomization ranges, with Bayesian Domain Randomization (BayRn), which adaptively optimizes distributions using Bayesian Optimization for more precise but computationally costly results.

• EKF and ROS : implementation of an Extended Kalman Filter applied to a single and double pendulum on ROS.

• RL-fundamentals : use of Gym Environment used to control the cartpole agent exploiting different reward functions.

• Q-learning : training cartpole with Q-Learning Temporal-Difference algorithm.

• Policy-gradient : use of Policy Gradient algorithms like PPO and SAC from Stable-Baselines3 library.

Spiking Neural Networks - group project :
Implemented Spiking Neural Networks (SNN) in Rust to simulate neural dynamics. Controlled errors were introduced into spike activity to evaluate robustness and fault tolerance.

Nonlinear Control and Aerospace Applications :
Implementation of various nonlinear control techniques applied to aerospace scenarios, including Feedback Linearization (FL), Sliding Mode Control (SMC) and Nonlinear Model Predictive Control (NMPC).

Modeling and Control of Cyber-Physical Systems - group projects :

• Sensor system modeling resilient to adversarial attacks using Iterative Shrinkage Thresholding Algorithm (ISTA).
• Control of a set of magnetic levitator dynamics exploiting a regulator and an observer, with a comparison of centralized and distributed approaches.

Technologies for Autonomous Vehicles - group projects :

• Image processing : implemented a machine learning algorithm to detect driver drowsiness and focus. Eye openness and gaze direction were analyzed using facial landmark points (Mediapipe library).
• Vehicle dynamics : a vehicle dynamics simulator implemented using the Pacejka tire model, with ABS and Traction Control Systems (TCS) using PID controllers. Tests were executed for maximum velocity and braking on different surfaces.
• Vehicle control : control of an autonomous vehicle to follow a trajectory. Various control techniques were applied and compared for performance.

Implementation of a GAN Applied to an Image Inpainting Task - group project :

Network Dynamics - group project :
a series of three homework completed during the course and focused on modeling agent systems dynamics applying Graph Theory with the help of NetworkX Python library.