[Paper Review] Build your own visual-inertial odometry aided cost-effective and open-source autonomous drone

Inkyu Sa, Mina Kamel+6 more|arXiv (Cornell University)|Aug 22, 2017

Robotics and Sensor-Based Localization22 references3 citations

TL;DR

This paper presents a low-cost, open-source VTOL drone platform using off-the-shelf visual-inertial sensors and an onboard computer, calibrated for visual-inertial odometry and model predictive control. It achieves 0.036m RMS pose error during hover and 0.82% drift over 180m flight, demonstrating robust performance in indoor and outdoor environments with wind disturbances.

ABSTRACT

This paper describes an approach to building a cost-effective and research grade visual-inertial odometry aided vertical taking-off and landing (VTOL) platform. We utilize an off-the-shelf visual-inertial sensor, an onboard computer, and a quadrotor platform that are factory-calibrated and mass-produced, thereby sharing similar hardware and sensor specifications (e.g., mass, dimensions, intrinsic and extrinsic of camera-IMU systems, and signal-to-noise ratio). We then perform a system calibration and identification enabling the use of our visual-inertial odometry, multi-sensor fusion, and model predictive control frameworks with the off-the-shelf products. This implies that we can partially avoid tedious parameter tuning procedures for building a full system. The complete system is extensively evaluated both indoors using a motion capture system and outdoors using a laser tracker while performing hover and step responses, and trajectory following tasks in the presence of external wind disturbances. We achieve root-mean-square (RMS) pose errors between a reference and actual trajectories of 0.036m, while performing hover. We also conduct relatively long distance flight (~180m) experiments on a farm site and achieve 0.82% drift error of the total distance flight. This paper conveys the insights we acquired about the platform and sensor module and returns to the community as open-source code with tutorial documentation.

Motivation & Objective

• To develop a cost-effective, research-grade VTOL drone platform using mass-produced, factory-calibrated hardware components.
• To enable accurate visual-inertial odometry and multi-sensor fusion without extensive parameter tuning through system calibration and identification.
• To validate the platform's performance under real-world conditions, including wind disturbances and long-distance flight.
• To share the complete system, including code and tutorials, as open-source for the research community.

Proposed method

• Utilizes off-the-shelf visual-inertial sensors, an onboard computer, and a quadrotor airframe with consistent hardware specifications.
• Performs system calibration and identification to align intrinsic and extrinsic parameters of the camera-IMU system.
• Employs visual-inertial odometry and model predictive control frameworks for state estimation and flight control.
• Uses motion capture systems indoors and laser trackers outdoors for ground truth trajectory validation.
• Applies multi-sensor fusion techniques to improve state estimation accuracy under dynamic conditions.
• Conducts extensive testing on hover, step responses, and trajectory following tasks in diverse environments.

Experimental results

Research questions

• RQ1 Can a low-cost, off-the-shelf hardware platform achieve high-precision visual-inertial odometry with minimal calibration effort?
• RQ2 How does the system perform in terms of pose accuracy and drift under external wind disturbances?
• RQ3 To what extent can factory-calibrated components reduce the need for manual parameter tuning in autonomous drone systems?
• RQ4 What level of trajectory tracking accuracy can be achieved over long-distance flight in real-world outdoor conditions?

Key findings

• The system achieves a root-mean-square (RMS) pose error of 0.036 meters during hover, demonstrating high accuracy in controlled indoor conditions.
• Over a 180-meter outdoor flight on a farm site, the platform exhibits only 0.82% drift relative to the total distance flown, indicating strong long-term stability.
• The use of factory-calibrated components significantly reduces the need for extensive parameter tuning, enabling faster system integration.
• The platform maintains robust performance under external wind disturbances, as validated through step response and trajectory following tasks.
• The open-source release of the code and tutorial documentation enables reproducibility and community extension of the platform.