High-Speed Stereo Visual SLAM for Low-Powered Computing Devices
Ashish Kumar, Jaesik Park, Laxmidhar Behera
Abstract
We present an accurate and GPU-accelerated Stereo Visual SLAM design called Jetson-SLAM. It exhibits frame-processing rates above <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\mathbf {60}$</tex-math></inline-formula> FPS on NVIDIA's low-powered <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\mathbf {10}$</tex-math></inline-formula> W Jetson-NX embedded computer and above <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\mathbf {200}$</tex-math></inline-formula> FPS on desktop-grade <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\mathbf {200}$</tex-math></inline-formula> W GPUs, even in stereo configuration and in the multiscale setting. Our contributions are threefold: ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i</i> ) a Bounded Rectification technique to prevent tagging many non-corner points as a corner in FAST detection, improving SLAM accuracy. ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ii</i> ) A novel Pyramidal Culling and Aggregation ( <monospace xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">PyCA</monospace> ) technique that yields robust features while suppressing redundant ones at high speeds by harnessing a GPU device. <monospace xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">PyCA</monospace> uses our new Multi-Location Per Thread culling strategy ( <monospace xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">MLPT</monospace> ) and Thread-Efficient Warp-Allocation ( <monospace xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">TEWA</monospace> ) scheme for GPU to enable Jetson-SLAM achieving high accuracy and speed on embedded devices. ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">iii</i> ) Jetson-SLAM library achieves resource efficiency by having a data-sharing mechanism. Our experiments on three challenging datasets: KITTI, EuRoC, and KAIST-VIO, and two highly accurate SLAM backends: Full-BA and ICE-BA show that Jetson-SLAM is the fastest available accurate and GPU-accelerated SLAM system (Fig. 1). <fig id="fig1" orientation="portrait" position="float" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <label>Fig. 1.</label> <caption> (a) Output of Jetson-SLAM's GPU-accelerated and resource-efficient Frontend–Middle-end design, (b) the output trajectory, (c) Frames-Per-Second benchmarking on Jetson-NX embedded computer, and (d) SLAM performance on a KITTI sequence. </caption> <graphic orientation="portrait" position="float" xlink:href="kumar1-3329621.eps"/></fig>