A Comparative Computational Study of Sensor Fusion Architectures, Motion Planning Algorithms, and Control Strategies for Autonomous Ground Vehicles and UAVs

Abstract

Khushvir Singh

This paper presents an original computational study comparing sensor fusion architectures, motion planning algorithms, and propulsion energy models for autonomous ground vehicles (AVs) and unmanned aerial vehicles (UAVs). Using physics-based simulation in Python, we evaluate four sensor fusion strategies — GPS-only, LiDAR-IMU, Camera-IMU, and Full Fusion — across five operational environments; four motion planning algorithms — A*, RRT*, D*, and Model Predictive Control (MPC) — across three map complexities; and three UAV attitude controllers — PID, MPC, and Adaptive — under varying wind disturbances. We further model propulsion energy consumption and operational range for lithium-ion and hydrogen fuel-cell powertrains across five representative mission profiles. Key findings: Full Fusion reduces localisation RMSE by up to 86% over GPS-only in urban canyons; MPC achieves the lowest path jerk (1.23 m/ s3) and highest success rate (93.7%) in complex scenarios; MPC attitude control reduces UAV roll RMSE by 83% versus PID under strong wind; and hydrogen fuel-cell propulsion delivers 2.3× greater range for AVs and up to 2.4× longer flight endurance for UAVs compared to lithium-ion at equivalent energy weight.

PDF