In distributed multi-agent systems, each agent solves a local optimization problem using its own state information and limited communication with neighboring vehicles, while collectively maintaining formation, avoiding collisions, and tracking shared mission objectives. The project will investigate scalable coordination strategies, robustness to disturbances and communication losses, and real-time implementation on representative swarm navigation and surveillance scenarios.
- Task 1: Drone swarm modelling
- Develop dynamic and kinematic models for a swarm of quadrotors, including inter-agent communication, collision-avoidance constraints, actuator limits, and external disturbances such as wind.
- Task 2: Distributed MPC algorithm
- Develop a distributed model predictive control algorithm in which each drone computes its trajectory using local measurements and information exchanged with neighboring drones. Establish conditions for feasibility, stability, and collision-free operation.
- Task 3: Simulation and benchmarking
- Implement the distributed MPC framework in a multi-drone simulation environment. Evaluate formation control, trajectory tracking, obstacle avoidance, scalability, and robustness to communication delays and packet losses.
- Task 4: Crazyflie platform development
- Develop the experimental platform using Crazyflie nano-quadrotors, including positioning, communication, state estimation, safety mechanisms, and interfaces between the distributed MPC software and the Crazyflie control stack.
- Task 5: Real-time implementation
- Implement and optimize the distributed MPC algorithm for real-time operation on a Crazyflie swarm. Investigate centralized, onboard, and edge-computing architectures to determine suitable computation and communication arrangements.
- Task 6: Experimental validation
- Validate the developed methods through Crazyflie experiments involving formation changes, coordinated trajectory tracking, obstacle avoidance, and recovery from communication or agent failures. Compare the results with conventional consensus and centralized MPC approaches.
- Task 7: Write the thesis
- Introduction, problem statement, literature review, methods, discussion, results.
Timeline
Master thesis projects can be taken in 10-20-30 credit formats (i.e., 10 credits in Semester 2, 20 credits in Semester 3, and 30 credits in Semester 4) or 30-30 credit (i.e., 30 credits in Semester 3 and 30 credits in Semester 4) formats. In my experience, the latter (30-30 credits) works better for students as you can devote your full attention to the thesis without coursework as a distraction. Nonetheless, here are proposed timelines for each system.
10-20-30 Credits
Semester 2
Begin Tasks 1, 2, and 3. Document results into a semester essay, along with the problem statement and literature review from Task 7. Begin Task 4 by familiarizing yourself with the CrazyFlie codebase and workflow.
Semester 3
Complete Tasks 1, 2, and 3. Document the results in the Methods section of Task 7. Continue Task 4, while beginning Tasks 5 and 6.
Semester 4
Complete Tasks 4, 5, and 6. Write the results and critically analyze them in the discussion of Task 7. Complete Task 7.
30-30 Credits
Semester 3
Begin and complete Tasks 1, 2, and 3. Document results into a semester essay, along with the problem statement and literature review from Task 7. Begin and complete Task 4 by familiarizing yourself with the CrazyFlie codebase and workflow. Document the results in the Methods section of Task 7. Begin Tasks 5 and 6.
Semester 4
Complete Tasks 5, and 6. Write the results and critically analyze them in the discussion of Task 7. Complete Task 7.