Topographic SLAM Using a Single Terrain Altimeter in GNSS-Restricted Environment

Junwoo Jang,Jinwhan Kim

doi:10.1109/access.2022.3145978

Abstract

In a Global Navigation Satellite System (GNSS)-restricted area, a mobile robot navigation system exploits surrounding environment information. For an aerial or underwater vehicle, undulating terrain of a land or seabed surface is a valuable information resource that leads to the development of terrain-referenced navigation (TRN) algorithms. However, due to the vast amount of a vehicle’s activity area, surveying all the regions to obtain a high-resolution terrain map is impractical and requires simultaneous localization and mapping (SLAM) as a highly desirable capability. This paper presents a topographic SLAM algorithm using only a single terrain altimeter, which is low-cost, computationally efficient, and sufficiently stable for long-term operation. The proposed rectangular panel map structure and update method enable robust and efficient SLAM. As terrain elevation changes are inherently nonlinear, an extended Kalman filter (EKF)-based SLAM filter is adopted. The feasibility and validity of the proposed algorithm are demonstrated through simulations using terrain elevation data from a real-world undersea environment.

Highlights

L OCALIZATION is one of the most critical capabilities for autonomous robotic vehicles. These vehicles are generally equipped with proprioceptive motion sensors for dead-reckoning using onboard motion sensors with no position fixes that are susceptible to integration drift error, which grows in time without bound and can significantly degrade navigation accuracy
This position estimate is calculated by combining both motion measurements provided by onboard motion sensors and position fixes obtained from the map that is being built
The multibeam echosounder (MBE) is a more commonly-used and effective topographic sensor for seabed terrain mapping, it cannot be integrated into a lost-cost small autonomous underwater vehicles (AUVs) due to its weight, power consumption, and computational cost

Summary

INTRODUCTION

L OCALIZATION is one of the most critical capabilities for autonomous robotic vehicles. The proposed bathymetric SLAM approach enables generating a bathymetric map using a series of depth measurements from a single-beam terrain altimeter while estimating the vehicle’s position along with onboard motion sensors simultaneously. This position estimate is calculated by combining both motion measurements provided by onboard motion sensors and position fixes obtained from the map that is being built. The motion model is often expressed as kinematic equations driven by inertial sensor measurements as the control input u, consisting of three accelerations and three angular rates In this problem formulation, no explicit position information is assumed to be provided to the SLAM filter as in the scenario where there are no GNSS signal available. The vehicle’s absolute (pressure) altitude relative to the reference surface is measured by proprioceptive sensors, such as barometers or pressure sensors measuring the static pressure of the ambient fluid (e.g., air or water)

PANEL-BASED MAP STRUCTURE

UPDATE IN EKF-SLAM FILTER

TOPOGRAPHIC SLAM SIMULATIONS

VEHICLE DYNAMICS FOR SIMULATIONS

CONCLUSIONS