A Hardware Approach for Solving the Robot Localization Problem Using a Sequential EKF

Abstract

This work describes a hardware architecture for implementing a sequential approach of the Extended Kalman Filter (EKF) that is suitable for mobile robotics tasks, such as self-localization, mapping and navigation problems. As such algorithm is computationally intensive, commonly it is implemented in PC-based platforms to be employed on larger robots. In order to allow the development of small robotic platforms, as those required in many current state of the art research (for instance microrobotics area), small size, low-power and high precision computation capabilities are required. This work proposes a hardware architecture for self-localization task, allowing the fusion of data coming from different sensors such as ultrasonic and laser rangefinder. The architecture is based on a single precision floating-point arithmetic, allowing the operations to be computed with large precision and dynamic range. The system has been adapted for achieving a reconfigurable platform, suitable for research tasks, and applied to a Pioneer 3AT mobile robot platform from Mobile Robots Inc. In order to compare the performance of the system, the same localization technique has been implemented in a high-end PC, as well as using an embedded microprocessor Nios II from Altera Inc. In this paper several metrics have been used to evaluate system performance and suitability, measuring FPGA resources consumption and performance. Finally, the suitability of reconfigurable devices for such kind of applications has been verified and also discussed.

Hardware and Software Co-design for the EKF Applied to the Mobile Robotics Localization Problem

“…is the nonlinear function of the measurement system, x K-1 is the state vector, u K-1 is the input (or control) vector and z k the output vector. For discrete time, the EKF prediction stage will be defined by (2) and (3). The EKF estimation stage is defined by (4), (5) and (6).…”

“…The algorithm implementation was accomplished considering the following approach: 1) a software implementation in the Nios II processor (written in C language) for the EKF Prediction Stage and 2) a hardware implementation [2] in the Altera Cyclone IV FPGA (written in VHDL code) for the EKF Estimation Stage. Fig.…”

“…To implement those operations, a Finite State Machine (FSM) [2] was designed based on floating-point arithmetic operators [9]. The FSM structure is shown in Fig.…”

“…Fig. 4 shows the data path [2] for implementing the estimation stage architecture. The EMB unit is composed of the arrays elements In the implementation of this EKF stage was necessary to divide in eleven steps as shown in (16) to (26).…”

Hardware Architecture of the EKF Prediction Stage applied to mobile robot localization

“…To accomplish this, a Hardware Module for the prediction stage was developed, which uses floating-point representation for the intensive matrix calculations of the Sequential EKF. This module was connected to another module corresponding to the estimation stage previously developed in [2], allowing the complete implementation of the EKF algorithm in hardware. The algorithm has been implemented on Altera Cyclone IV FPGA with Nios II Processor and tested over a P3AT mobile platform [4].…”