Run Your Visual-Inertial Odometry on NVIDIA Jetson: Benchmark Tests on a Micro Aerial Vehicle

Abstract

This paper presents benchmark tests of various visual(-inertial) odometry algorithms on NVIDIA Jetson platforms. The compared algorithms include mono and stereo, covering Visual Odometry (VO) and Visual-Inertial Odometry (VIO): VINS-Mono, VINS-Fusion, Kimera, ALVIO, Stereo-MSCKF, ORB-SLAM2 stereo, and ROVIO. As these methods are mainly used for unmanned aerial vehicles (UAVs), they must perform well in situations where the size of the processing board and weight is limited. Jetson boards released by NVIDIA satisfy these constraints as they have a sufficiently powerful central processing unit (CPU) and graphics processing unit (GPU) for image processing. However, in existing studies, the performance of Jetson boards as a processing platform for executing VO/VIO has not been compared extensively in terms of the usage of computing resources and accuracy. Therefore, this study compares representative VO/VIO algorithms on several NVIDIA Jetson platforms, namely NVIDIA Jetson TX2, Xavier NX, and AGX Xavier, and introduces a novel dataset 'KAIST VIO dataset' for UAVs. Including pure rotations, the dataset has several geometric trajectories that are harsh to visual(inertial) state estimation. The evaluation is performed in terms of the accuracy of estimated odometry, CPU usage, and memory usage on various Jetson boards, algorithms, and trajectories. We present the results of the comprehensive benchmark test and release the dataset for the computer vision and robotics applications.

Toward Efficient and Robust Multiple Camera Visual-inertial Odometry

“…Although these VIO methods have achieved good results for online state estimation, The CPU usages are all high because massive visual images need to be processed. According to the experiments conducted on Xavier [4], the CPU usage of VINS-Mono is about 150-170% on multicore processing, MSCKF is above 170%. ROVIO has the least CPU usages (around 60%), but the back-end EKF is less efficient compared to the bundle adjustment method [13].…”

“…Consequently, it takes a large proportion of CPU resources to guarantee real-time performance. Jeon [4] tests the CPU usage of different VIO algorithms on various NVIDIA hardware (Jetson TX2, Xavier NX, and AGX Xavier boards). His work shows that Fig.…”