Cost-Based MPPI: Enhancing the Efficiency of MPPI Controllers in 3D Space for UAV Control

In this paper, we explored an improved model predictive path integral (MPPI) controller for trajectory tracking of Unmanned Aerial Vehicles (UAVs). The proposed improved MPPI introduces adaptive control parameters (the horizon step T and the sample rollouts K) to enhance the sample efficiency of the traditional MPPI. As a Monte Carlo-based optimal controller, the MPPI guarantees its performance only when the number of samples is large. To improve the sample efficiency of the existing MPPI, we introduced adaptive control parameters to develop an improved MPPI that can effectively generate trajectories even with a small number of samples. The adaptation rules for the control parameters are calculated based on real-time cost function evaluations. This adaptive strategy aims to provide MPPI trajectories more efficiently, promoting more stable and efficient UAV flight paths. The validity and feasibility of the proposed algorithm were evaluated through UAV flight trajectory tracking simulations.