ORB-SLAM accelerated on heterogeneous parallel architectures

Abstract

SLAM algorithm permits the robot to cartography the desired environment while positioning it in space. It is a more efficient system and more accredited by autonomous vehicle navigation and robotic application in the ongoing research. Except it did not adopt any complete end-to-end hardware implementation yet. Our work aims to a hardware/software optimization of an expensive computational time functional block of monocular ORB-SLAM2. Through this, we attempt to implement the proposed optimization in FPGA-based heterogeneous embedded architecture that shows attractive results. Toward this, we adopt a comparative study with other heterogeneous architecture including powerful embedded GPGPU (NVIDIA Tegra TX1) and high-end GPU (NVIDIA GeForce 920MX). The implementation is achieved using high-level synthesis-based OpenCL for FPGA and CUDA for NVIDIA targeted boards.

AFO-SLAM: an improved visual SLAM in dynamic scenes using acceleration of feature extraction and object detection

“…In recent years, there has been a shift in research towards identifying suitable architectures for real-time operation of end-to-end embedded SLAM systems [40]. As a result, researchers have turned to heterogeneous architectures to fully exploit the performance advantages of powerful devices like GPUs [41]. Several works have explored the use of GPU to accelerate parallel calculation of certain algorithms in SLAM.…”

“…Jeon et al [40] compared different acceleration methods for ORB-SLAM and visual odometry system on different platforms using NVIDIA GPU. Mamri et al [41] implemented parallelization using OpenCL and CUDA on FPGA and NVIDIA platforms to improve system performance by optimizing timeconsuming modules. Aldegheri et al [13] demonstrated significant performance improvements for the acceleration of ORB-SLAM2 on NVIDIA platforms using CUDA and OpenCV.…”

Brief Announcement: Optimized GPU-accelerated Feature Extraction for ORB-SLAM Systems

“…ORB-SLAM features a considerable computational time so some works tried to reduce this by exploiting accelerators. In [10], a variety of acceleration methods for ORB-SLAM or more generic Visual Odometry, are compared; in [13], the authors accelerated the method using OpenCL for FPGA and NVIDIA platforms; in [12,1], CUDA is exploited to accelerate ORB-SLAM2 on NVIDIA platforms using CUDA and OpenCV 1 to extract and match features in Stereo camera systems. Moreover, in [1], the authors use OpenVX to offload computation to GPU.…”