Beyond Reference Trajectories: A Waypoint-Based Model Predictive Path Integral Control for Agile Drone Racing
Fangguo Zhao, Xin Guan, Shuo Li
AI summary
Problem
Conventional model-based drone racing controllers rely on pre-computed reference trajectories, limiting flexibility and generalization, while reinforcement learning methods lack transferability and require extensive training. This work addresses how to integrate RL-inspired gate progress objectives into model-based control without relying on reference paths.
Approach
The authors implement a sampling-based MPPI controller that directly optimizes a discontinuous gate-progress objective using only waypoint positions, bypassing the need for reference trajectories. They also build a unified simulation testbed to systematically compare this reference-free method against classical trajectory tracking and contouring control objectives.
Key results
- Novel reference-free MPPI controller using a gate progress objective
- Unified MPPI-based testbed for fair comparison of control objectives
- Reference-free gate progress achieves near time-optimal racing performance
- MPPI implementations of traditional objectives match or surpass gradient-based solvers
Why it matters
It enables robust, generalizable, time-optimal drone racing without offline trajectory planning, bridging the gap between reinforcement learning reward shaping and model-based control for agile robotics.
Abstract
While model-based controllers have demonstrated remarkable performance in autonomous drone racing, their performance is often constrained by the reliance on pre- computed reference trajectories. Conventional approaches, such as trajectory tracking, demand a dynamically feasible, full-state reference, whereas contouring control relaxes this requirement to a geometric path but still necessitates a reference. Recent advancements in reinforcement learning (RL) have revealed that many model-based controllers optimize surrogate objec- tives, such as trajectory tracking, rather than the primary racing goal of directly maximizing progress through gates. Inspired by these findings, this work introduces a reference- free method for time-optimal racing by incorporating this gate progress objective, derived from RL reward shaping, directly into the Model Predictive Path Integral (MPPI) formulation, which only depends on waypoint positions. The sampling- based nature of MPPI makes it uniquely capable of optimizing the discontinuous and non-differentiable objective in real-time. We also establish an empirical testbed that leverages MPPI to systematically and fairly compare three distinct objective functions with a consistent dynamics model and parameter set: classical trajectory tracking, contouring control, and the proposed gate progress objective. We compare the performance of these three objectives when solved via both MPPI and a tra- ditional gradient-based solver. Our results demonstrate that the proposed reference-free approach achieves competitive racing performance, rivaling or exceeding reference-based methods.