• Software-in-the-loop simulation of a group of autonomous drones for dynamic coverage of a partially known area :
  This thesis focuses on aerial collaborative robotics, particularly on trajectory planning and fleet management for drones operating in applications such as traffic monitoring, urban delivery, and infrastructure inspection. The goal is to develop a simulation framework in ROS2/Gazebo to validate a fleet management system in partially known environments. The student will be required to implement an optimization protocol to efficiently generate drone trajectories, taking into account factors such as communication range, agent dynamics, and sensor models. The work will include a critical analysis of the results and an evaluation of the mission feasibility for potential experimental implementation.

Required Skills:

• Familiarity with ROS2 and Gazebo for simulation development.

• Good programming skills in Python and/or C++ for implementing ROS2 nodes controlling autonomous drones.

• Basic knowledge of automatic control theory is helpful for understanding theoretical concepts

• Software-in-the-loop simulation of a group of autonomous robots in a formation scheme with time delays in the communication network:
  Formation control is one of the primary tasks in managing a group of robots, as it helps avoid collisions and maintain a desired safety distance. The presence of communication delays within the network poses a significant risk to formation control, especially if these delays are not properly considered. For this reason, it is essential to include the effect of communication delays among agents in any realistic simulation. This simulation-based thesis aims to develop a simulation framework in a ROS2/Gazebo environment to analyze the impact of time delays on a formation control algorithm for a group of ground robots. The student will be required to implement a consensus protocol to generate the formation in a decentralized manner and to appropriately integrate time delays into the simulation. Additionally, a critical analysis of how the critical delay thresholds for stability depend on the physical and control parameters of the system is required.

Required Skills:

• • Familiarity with ROS2 and Gazebo for simulation development.

  • Good programming skills in Python and/or C++ for implementing ROS2 nodes controlling autonomous drones.

  • Basic knowledge of automatic control theory is helpful for understanding theoretical concepts

• Integration of Reinforcement Learning and LTL-Based Reward Machines for Event-Triggering Consensus in Multi-Robot Systems: Formation and Decentralized Estimation:
  The use of multi-robot systems is continuously expanding, particularly in applications such as logistics, exploration, and delivery. To cooperate in complex environments, consensus theory can be employed to achieve a common goal in a decentralized way. Since the network must share information while optimizing the use of communication resources, it is crucial to strategically decide when to trigger communication to update the shared state. Reinforcement Learning (RL) provides a flexible mechanism for learning communication activation policies based on local measurements. However, traditional reward shaping in RL does not always guarantee compliance with high-level general constraints. A Reward Machine (RM), derived from Linear Temporal Logic (LTL) specifications, can translate the rules to be followed into an interpretable structure and assigns coherent rewards related to the achievement of the objectives. The thesis focuses on two prominent multi-agent applications: formation and decentralized estimation. In these scenarios, the master student is expected to develop: 

  • A consensus module to synchronize the state variables depending on the specific application (e.g. relative positions or target estimates).

  • A communication module, controlled by an RL agent, to determine when to exchange information about each agent’s local state to update the consensus process.

  • A Reward Machine (RM), formulated based on LTL principles, to evaluate the effectiveness of the communication events, assigning rewards or penalties depending on whether, after communication, the formation approaches the desired triangular layout. 

Required Skills:

• • Familiarity with Matlab/Simulink for simulation development.

  • Good programming skills in Python and/or C++ for implementing ROS2 nodes controlling autonomous drones.

  • Good knowledge of automatic control theory is helpful for understanding theoretical concepts

• Orbital simulator for rendezvous and proximity operations with integrated robotic-arm MPC

Recent interest in on-orbit servicing, debris removal, asteroid resource utilization, and close-proximity inspection makes reliable rendezvous and interaction with non-cooperative objects a critical capability. Proximity operations require tight coordination between spacecraft guidance, attitude control, and any manipulator used for capture or servicing. MPC offers an attractive control framework because it handles constraints explicitly and can optimize multivariable behavior over a prediction horizon, which is important when coordinating arm motion with spacecraft dynamics and collision-avoidance constraints. 

This thesis will develop a high-fidelity orbital simulator tailored to rendezvous and proximity operations (RPO) with uncooperative targets (satellites, debris or small asteroids). The project extends the orbital simulator by integrating a kinematic/dynamic model of a manipulator and implements a Model Predictive Control (MPC) to coordinate spacecraft guidance and robotic-arm motions during close-proximity interactions. The work will be implemented in the MATLAB/Simulink environment and validated through representative scenarios. 

Required skills:

• Proficiency with MATLAB/Simulink

• Strong theoretical background in control (MPC)

• Solid understanding of orbital mechanics and relative motion dynamics