Spiral-Based Reactive Obstacle Avoidance for UAVs with PX4 SITL Integration

The noticeable rise of drones, or UAVs, in business, rescue operations, and general ap- plications necessitates increased safety due to their growing presence. As drones begin to operate alongside humans and existing structures, there is a need for preventing collisions with objects and/or people. This thesis presents a potential solution to the obstacle avoidance problem, enabling drones to avoid possible collisions during navigation. The algorithm relies on an Archimedean spiral to identify an escape point in real time based on environmental data perceived by the depth Camera. During navigation, the algorithm continuously processes the point cloud information, generating a suitable escape point if an obstruc- tion is detected within a safety cylindrical volume. The implementation leverages the co-simulation between ROS2 and PX4 Autopilot via uORB messages for seamless communication between high-level setpoint tracking and low-level flight controllers. Octomap processes the point cloud provided by the depth camera to retrieve a probabilistic 3D representation of the surrounding envi- ronment while maintaining reduced memory occupancy. Extensive Gazebo simulation validates system effectiveness across different scenarios.