Design And Evaluation of a Lightweight Multi-Sensor Fusion Visual Navigation System for Uavs

Yunyi Song; Xuyang Sun

doi:10.54097/acsmyx32

Authors

Yunyi Song
Xuyang Sun

DOI:

https://doi.org/10.54097/acsmyx32

Keywords:

UAV; Visual–inertial odometry; multi-sensor fusion; Error analysis; Trajectory evaluation.

Abstract

With the development of visual–inertial navigation technology, this paper evaluates the deployment performance of visual–inertial odometry (VIO) algorithms on lightweight UAV platforms. Three representative algorithms, MSCKF, VINS-Mono, and OpenVINS, are selected and tested on the EuRoC MH_01_easy dataset under a same virtual platform (ROS Noetic + Ubuntu 20.04, 4-core CPU, 8GB RAM). Absolute Trajectory Error (ATE) and Relative Pose Error (RPE) are used as the main evaluation metrics. The results show that VINS-Mono achieves the highest accuracy, with a mean ATE of 0.078 m and RPE ranging from 0.18 to 0.33 m. MSCKF presents slightly lower accuracy (mean ATE of 0.090 m) but has the lowest CPU and memory usage with strong stability. OpenVINS reaches accuracy close to VINS-Mono (mean ATE of 0.080 m), with moderate resource consumption but higher sensitivity during initialization. Resource monitoring shows that the computational cost of VINS-Mono is about 1.8 times that of MSCKF. These findings highlight the trade-offs among accuracy, computational efficiency, and deployment adaptability, providing practical references for low-altitude applications such as urban logistics and power line inspection.

References

[1]Zhang T, Li Q, Zhang C, et al. Development trends of intelligent unmanned autonomous systems. Unmanned Systems Technology, 2018, 1(1): 11–22.

[2]Zou J, Sun Q, Liu W, et al. An improved MSCKF visual–inertial odometry method for indoor scenes. Unmanned Systems Technology, 2024, 7(3): 54–66.

[3]Li X, Zhang G, Xie B. An improved visual odometry method based on VINS. Foreign Electronic Measurement Technology, 2023, 42(1): 20–27.

[4]Huang S, Wei G, Guan Q, et al. A positioning method based on fusion of IMU arrays and camera data. Mini-micro Systems, 2022, 43(12): 2620–2627.

[5]Shan T, Englot B. LeGO-LOAM: Lightweight and ground-optimized lidar odometry and mapping on variable terrain. In: Proc. IEEE/RSJ Int. Conf. Intelligent Robots and Systems (IROS), Madrid, Spain, Oct. 1–5, 2018.

[6]Zeng C, Zheng Q, Cheng S. Indoor robot localization system based on VINS-Mono. Transducer and Microsystem Technologies, 2022, 41(11): 85–88.

[7]Wang Z H, Liang D T, Liang D, et al. A SLAM method based on inertial/magnetic sensors and monocular vision fusion. Robot, 2018, 40(6): 933–941.

[8]Wang P, Li N, Zhao W. Energy consumption optimization and experimental verification of lightweight UAV visual–inertial navigation systems. Acta Aeronautica et Astronautica Sinica, 2023, 44(5): 320–332.

[9]Qin T, Li P, Shen S. VINS-Mono: A robust and versatile monocular visual–inertial state estimator. IEEE Transactions on Robotics, 2018, 34(4): 1004–1020.

[10]Geneva P, Eckenhoff K, Huang G, et al. OpenVINS: A research platform for visual–inertial estimation. IEEE Robotics and Automation Letters, 2020, 5(2): 1422–1429.