ÄûÃʵ¼º½

The title of the thesis: Towards Beyond Visual Line Of Sight (BVLOS) Risk-Aware UAV Path Planning: Ensuring Swarm Safety, Obstacle Avoidance, and Ground Crash Risk Mitigation

Thesis defender: Sihem Ouahouah
Opponent: Prof. Heidi Kuusniemi, Tampere University, Finland
Custos: Prof. Riku Jäntti, Aalto University School of Electrical Engineering

After being restricted for decades to military use, Unmanned Aerial Vehicles (UAVs) have recently attracted significant interest from both academia and industry due to their compact sizes, ability to carry onboard payloads, high mobility, and strong connectivity. UAVs have been proposed for a wide range of new, multidisciplinary, and emerging applications. Realizing these use cases, however, requires a transition from Visual Line of Sight (VLOS) operation to Beyond Visual Line of Sight (BVLOS) operation. The reliance on VLOS is mainly due to UAV navigation regulations, limited network communication ranges, and UAVs’ energy constraints.

Fortunately, recent advances offer promising opportunities. On one side, 5G networks provide Ultra-Reliable Low-Latency Communications (URLLC) for UAVs, which can help overcome many of the communication challenges associated with BVLOS operations. On the other side, aviation navigation authorities are gradually relaxing BVLOS regulations. Nevertheless, under these new circumstances, transitioning UAV navigation to BVLOS mode requires robust path-planning strategies that consider UAV limitations as well as global safety risks.

This research work focuses on improving the efficiency of UAV navigation management in BVLOS mode across the following contexts:

• Swarm UAV path planning in safe environments,where real-time path planning for UAV swarms is modeled using convex linear programming to ensure collision avoidance among swarm members while minimizing overall energy consumption.
• Autonomous UAV path planning within unsafe environments, where energy-efficient autonomous navigation systems for UAVs are enabled using reinforcement learning methods while ensuring collision avoidance with both static and mobile obstacles.
• Autonomous UAV path planning with ground risk mitigation, where 5G-exposed user-mobility data is leveraged to enable reinforcement-learning-based path planning that reduces the risk of UAV crashes in populated areas.

The results highlight the methods and considerations necessary for efficient autonomous UAV navigation. For each context, the thesis proposes new system architectures, algorithms, and performance results that demonstrate the feasibility of enabling UAV navigation in BVLOS mode.

The contributions of this work can support emerging applications such as urban transportation, network assistance, and agricultural irrigation.

Key words: Unmanned Aerial Vehicles (UAVs), Path planning, Risk mitigation

Thesis available for public display 7 days prior to the defence at .