DroneRanger: Simulation-Guided UAV Framework for Autonomous Wildlife Deterrence: Bridging PX4, ROS 2, and Unity through a Simulation-Driven Control Architecture that Enables Autonomous Navigation And Obstacle-Aware Wildlife Deterrence

Urban wildlife management poses increasing challenges in the Bay Area, where animals frequently damage vegetation, attract predators, and introduce safety hazards. Traditional control methods—such as chemical repellents, physical barriers, or mechanical traps—are costly, environmentally disruptive, and often ineffective. To address this, our project explores the use of a lightweight autonomous aerial platform to deter wildlife through non-contact cues, including motion, light, and short acoustic signals, thereby minimizing ecological impact. Building on this motivation, this paper presents the design and implementation of a drone obstacle detection and avoidance architecture that integrates perception and control into a unified simulation environment. The proposed architecture leverages the PX4 flight stack, ROS 2 middleware, and Gazebo/Unity environments to create a reproducible and extensible testing platform. For real-world validation, we employ the Holybro X500 V2 platform equipped with a Pixhawk 6C (PX4 Dev Kit). Synthetic depth camera data generated in simulation is used to evaluate feasibility and to ensure that the perception topics, autopilot interface, and control logic remain stable under realistic conditions. The results demonstrate that the proposed architecture enables reliable simulation-to-reality transfer for trajectory planning and obstacle avoidance in autonomous drone applications. This work establishes a foundation for deploying drones in ecologically sensitive contexts, offering a non-invasive, scalable, and safe approach to urban wildlife deterrence.