Preliminary Hardware and Software Design

Drone Design
Currently the hardware or software of the drone does not exist, however preliminary ideas have been formulated. This should sereve as a good baseline for what the drone needs to include, how it communicated and what safety features it should include. 
 Drone Subsystem Architecture 
 The drone is a custom-built, high-performance four rotor platform optimized for autonomous indoor navigation within a 10 x 10 x 4m netted flight cage. It features a decoupled, two-tier processing architecture: a dedicated Flight Controller (FC) handles real-time attitude estimation and low-level flight stability, while an onboard Companion Computer executes high-level vision processing, obstacle avoidance, and mission state machines. 
 System Specifications 
 Parameter Specification Engineering Note Total Mass 1.95 kg All-Up Weight (AUW) including 6S 4500mAh LiPo & Jetson Nano. Hover Power 380 W Average draw during steady hover, including onboard compute payload. Max Power 920 W Peak current draw during emergency maneuvers / dynamic correction. Power Source 6S 4500 mAh LiPo Sized to complete the 3-mission task cycle within the 30-min window. Primary Link 5.8 GHz (Wi-Fi) IEEE 802.11 compliant link for high-throughput video & telemetry. Safety Link 2.4 GHz Low-latency RadioMaster ExpressLRS manual safety link. 
 Hardware Integration & Component Breakdown 
 The hardware selection prioritizes structural rigidity, vibration isolation, and indoor localization reliability without reliance on satellite positioning. 
 Carbon Fiber Frame: Built on a 7-inch carbon fiber racing frame selected for high torsional stiffness and ample top/bottom deck area to accommodate the companion computer, power distribution, and downward-facing optical sensors. Propulsion System: Driven by four 1350 Kv brushless motors paired with 7-inch dual-blade propellers. This combination balances high thrust-to-weight ratio with electrical efficiency during steady hovering. Enclosure & Sensor Mounts: A custom 3D-printed housing protects sensitive core electronics from mechanical impacts and provides a rigid down-facing mount for the primary visual landing camera. Flight Control Stack: Uses a T-Motor Velox F7 SE Flight Controller combined with a V70A SE 4-in-1 ESC. The F7 FC is flashed with open-source PX4 or ArduPilot firmware to support MAVLink offboard commands. Indoor Positioning Sensors: Equipped with a Mateksys Optical Flow & LiDAR sensor package. This provides real-time ground velocity tracking and precise altitude measurement to replace GPS in enclosed, signal-degraded environments. Companion Computer & Vision Sensor: An onboard NVIDIA Jetson Nano processes high-level autonomy algorithms. It interfaces with a Global Shutter USB camera targeting ArUco markers. 
 Software & Autonomy Architecture 
 The high-level autonomy suite is developed and executed within MATLAB/Simulink . 
 Vision Processing Pipeline: A dedicated Simulink model performs real-time image processing in two stages: Target Probe Detection: Color-space thresholding (HSV) tuned specifically for yellow-green probe identification. Landing & Alignment: ArUco tag identification libraries compute relative 3D pose vectors for accurate landing alignment. Telemetry & Offboard Communications: High-level trajectory setpoints and real-time state estimates are streamed over MAVLink protocol via a 5.8 GHz Wi-Fi bridge. 
 Design Rationale: Global Shutter vs. Rolling Shutter in Indoor Vision: 
 Standard CMOS cameras use rolling shutters that expose pixels line-by-line. Under high drone vibration or swift pitch/roll maneuvers, this introduces severe image warping ("jello effect"), causing ArUco marker detection algorithms to fail or miscalculate target distance. 
 Integrating a Global Shutter camera captures the entire sensor plane simultaneously. This eliminates rolling-shutter distortion during motion, ensuring solid target tracking and accurate 3D pose estimation even during aggressive flight corrections. 
 Safety & Fail-Safe Architecture 
 To guarantee total flight control and satisfy safety protocols within the cage environment, the drone incorporates a triple-redundant fail-safe architecture. 
 Manual RC Override (Primary Safety): A dedicated physical switch assigned on the pilot’s RadioMaster Boxer transmitter immediately revokes offboard MATLAB control, returning raw flight control to the operator. Hard-Coded 3D Geofence (Boundary Protection): A cylindrical 3D geofence is defined directly inside the FC firmware, preventing the drone from drifting into the cage netting regardless of automated setpoint commands. GCS E-Stop (Ground Control Safety): The MATLAB/Simulink operator dashboard features an emergency software kill-switch that transmits an instant motor-disarm instruction. 
 Signal Loss Procedure (Link-Loss Recovery): 
 If the primary 5.8 GHz Wi-Fi data link or 2.4 GHz ExpressLRS control link experiences a complete disconnect, the F7 FC and T-Motor ESCs bypass state-estimation and execute an immediate, autonomous Land-In-Place sequence. The FC maintains attitude lock while decreasing throttle at a fixed, controlled descent rate to land vertically without drifting.