Docking two multirotors in midair using relative vision measurements

TL;DR

This paper introduces a relative state estimator and control system for autonomous midair docking of multirotors using onboard sensors, validated through experiments achieving centimeter-level precision.

Contribution

It presents a novel EKF-based state estimator and control architecture enabling precise midair docking using only onboard inertial and vision sensors.

Findings

Successful and repeatable midair docking demonstrated

Achieved relative position accuracy on the order of centimeters

Validated estimator and control system through experimental tests

Abstract

Modular robots have been rising in popularity for a variety of applications, and autonomous midair docking is a necessary task for real world deployment of these robots. We present a state estimator based on the extended Kalman filter for relative localization of one multirotor with respect to another using only onboard sensors, specifically an inertial measurement unit and a camera-marker pair. Acceleration and angular velocity measurements along with relative pose measurements from a camera on the first multirotor looking at a marker on the second multirotor are used to estimate the relative position and velocity of the first multirotor with respect to the second, and the absolute attitude of the first multirotor. We also present a control architecture to use these onboard state estimates to control the first multirotor at a desired setpoint with respect to the second. The performance of the estimator and control architecture are experimentally validated by successfully and repeatably performing midair docking -- a task that requires relative position precision on the order of a centimeter.