Strategizing at Speed: A Learned Model Predictive Game for Multi-Agent Drone Racing

TL;DR

This paper introduces a Learned Model Predictive Game approach for multi-agent drone racing, balancing strategic depth and computational speed, and demonstrates its superior performance over traditional methods in simulations and real-world tests.

Contribution

The paper proposes a novel Learned Model Predictive Game method that reduces latency in strategic planning for drone racing, outperforming existing MPG and MPC approaches.

Findings

LMPG outperforms MPG and MPC in head-to-head drone races.

MPG performs better at moderate speeds but struggles at high speeds due to latency.

LMPG achieves real-time strategic planning with improved race outcomes.

Abstract

Autonomous drone racing pushes the boundaries of high-speed motion planning and multi-agent strategic decision-making. Success in this domain requires drones not only to navigate at their limits but also to anticipate and counteract competitors' actions. In this paper, we study a fundamental question that arises in this domain: how deeply should an agent strategize before taking an action? To this end, we compare two planning paradigms: the Model Predictive Game (MPG), which finds interaction-aware strategies at the expense of longer computation times, and contouring Model Predictive Control (MPC), which computes strategies rapidly but does not reason about interactions. We perform extensive experiments to study this trade-off, revealing that MPG outperforms MPC at moderate velocities but loses its advantage at higher speeds due to latency. To address this shortcoming, we propose a Learned Model Predictive Game (LMPG) approach that amortizes model predictive gameplay to reduce latency. In both simulation and hardware experiments, we benchmark our approach against MPG and MPC in head-to-head races, finding that LMPG outperforms both baselines.