Efficient EKF-SLAM’s Jacobian Matrices Hardware Architecture and its FPGA Implementation

Abstract

Simultaneous Localization And Mapping (SLAM) algorithm allows a mobile robot to localize itself and move safely in an unknown environment while exploring and mapping it. The increasing use of SLAM in autonomous robots leads to the design of novel real-time hardware architectures. This paper presents an efficient EKF-SLAM’s (Extended Kalman Filter SLAM) Jacobian matrices hardware architecture design and implementation on a Field-Programmable Gate Array (FPGA). The proposed architecture is designed by using the Taylor Expansions with fixed-point arithmetic operators. The implementations for trigonometric functions cover the whole dynamic range i.e. [-π,π]. The proposed technique leads to higher accuracy and lower complexity. The design has been implemented on Cyclone II FPGA. It can reach up to 112 MHz and it uses 1517 logic elements, 1438 registers, and 44 embedded 9-bit multipliers, with 12 cycles of latency. Finally, a hardware resource comparison between this design and different state of the art techniques has been done. The proposed design uses very low resources with high frequency.