A novel edge-enabled SLAM solution using projected depth image information

  • PDF / 1,554,471 Bytes
  • 13 Pages / 595.276 x 790.866 pts Page_size
  • 1 Downloads / 229 Views

DOWNLOAD

REPORT


(0123456789().,-volV)(0123456789(). ,- volV)

MULTI-ACCESS EDGE COMPUTING ENABLED INTERNET OF THINGS

A novel edge-enabled SLAM solution using projected depth image information Jian-qiang Li1 Huihui Wang3



Yi-fan Zhang1 • Zhuang-zhuang Chen1 • Jia Wang1 • Min Fang2 • Cheng-wen Luo1



Received: 28 December 2018 / Accepted: 15 March 2019  Springer-Verlag London Ltd., part of Springer Nature 2019

Abstract Environmental mapping is the key step for mobile robots to perform tasks independently and perfectly. In recent years, visual SLAM, laser-based SLAM and simultaneous localization and mapping (SLAM) have aroused the interest of many people. Unfortunately, those technologies are not widely used, limited by the computational complexity, data processing and very low and predictable latency. This paper had mainly completed the following work and edge-enabled computingbased edge computing (Shi and Dustdar in Computer 49(5):78–81, 2016) is used as a solution to accelerate calculation. First of all, this research design works with inertial unit mobile robotic navigation systems, and all sensors are connected in edge layers in the framework of edge computing and explore the accelerometer, electronic compass, and gyroscope data. The accelerometer data are integrated using the Kalman filter data fusion algorithm to filter the random drift error caused by the gyroscope and the electronic compass. The state of the machine is determined by calculation of the corresponding attitude angle and position information. Second, a low-cost distance sensor is used to detect the depth and upload to the other fog node for computation. Next, the 3D point coordinate information is projected onto the two-dimensional coordinate extraction feature point to establish the feature map. Third, the extended Kalman filter SLAM is used to achieve simultaneous positioning and mapping. Finally, the method is validated in the experiment, proving that the method is feasible. The main improvement in this article is as follows: First, the multi-sensor data fusion algorithm is used to reduce the positioning error. Second, we use low-cost distance sensors to measure the depth of the model environment and reduce the cost. Third, we would take advantage of translating the three-dimensional depth information into a flat two-dimensional projection information to reduce the calculation of load and computing time. Fourth, our computation is distributed in different layers and focuses on edge-enabled platform to decrease the latency and redundancy. Keywords Mobile robots  Edge computing  Simultaneous localization and mapping  Extended Kalman filter  Inertial navigation  Depth sensor  Edge layer

& Jian-qiang Li [email protected] Yi-fan Zhang [email protected] Zhuang-zhuang Chen [email protected] Jia Wang [email protected]

Huihui Wang [email protected] 1

Department of Computer and Software Engineering, Shenzhen University, Shenzhen 518060, China

2

Department of Engineering, Harbin Institute of Technology, Harbin 150000, China