Project Title: SITL Model Generation from Flight Data Candidate: Som Shegokar Project Size: Large (350 hours) Location: Pimpri-Chinchwad, India
Abstract: High-fidelity simulation is critical for safe drone development, yet creating accurate SITL models is currently a manual process. My project aims to automate this by extracting dynamics directly from real-world ArduPilot flight logs (.bin files). I will develop a C++ tool to parse telemetry—including motor outputs, IMU data, and GPS velocities—to implement a “System Identification” pipeline.
Technical Approach: * Log Parsing: Extract high-frequency data (RCOU, IMU, ATT) using ArduPilot libraries. * System Identification: Calculate Thrust-to-Weight ratios, Moments of Inertia, and Drag Coefficients. * Model Generation: Automatically output standardized SITL JSON files and parameter configurations.
Timeline Highlights: * Bonding: Set up local SITL environment and fix a “Good First Issue” on GitHub. * Phase 1: Develop the .bin log parser and core estimation logic. * Phase 2: Implement complex dynamics (battery sag, rotational inertia) and build the final CLI tool.
I have already submitted my formal proposal on the GSoC portal and am eager to receive feedback from the mentors.