Pose estimation and data fusion algorithms for an autonomous mobile robot based on vision and IMU in an indoor environment

Abstract

Autonomous mobile robots became an active research direction during the past few years, and they are emerging in different sectors such as companies, industries, hospital, institutions, agriculture and homes to improve services and daily activities. Due to technology advancement, the demand for mobile robot has increased due to the task they perform and services they render such as carrying heavy objects, monitoring, delivering of goods, search and rescue missions, performing dangerous tasks in places like underground mines. Instead of workers being exposed to hazardous chemicals or environments that could affect health and put lives at risk, humans are being replaced with mobile robot services. It is with these concerns that the enhancement of mobile robot operation is necessary, and the process is assisted through sensors. Sensors are used as instrument to collect data or information that aids the robot to navigate and localise in its environment. Each sensor type has inherent strengths and weaknesses, therefore inappropriate combination of sensors could result into high cost of sensor deployment with low performance.

Regardless, the potential and prospect of autonomous mobile robot, they are yet to attain optimal performance, this is because of integral challenges they are faced with most especially localisation. Localisation is one the fundamental issues encountered in mobile robot which demands attention and the challenging part is estimating the robot position and orientation of which this information can be acquired from sensors and other relevant systems. To tackle the issue of localisation, a good technique should be proposed to deal with errors, downgrading factors, improper measurement and estimations. Different approaches are recommended in estimating the position of a mobile robot. Some studies estimated the trajectory of the mobile robot and indoor scene reconstruction using a monocular visual odmometry. This approach cannot be feasible for large zone and complex environment. Radio frequency identification (RFID) technology on the other hand provides accuracy and robustness, but the method depend on the distance between the tags, and the distance between the tags and the reader. To increase the localisation accuracy, the number of RFID tags per unit area has to be increased. Therefore, this technique may not result in economical and easily scalable solution because of the increasing number of required tags and the associated cost of their deployment. Global Positioning System (GPS) is another approach that offers proved results in most scenarios, however, indoor localization is one of the settings in which GPS cannot be used because the signal strength is not reliable inside a building. Most approaches are not able to precisely localise autonomous mobile robot even with the high cost of equipment and complex implementation. Most the devices and sensors either requires additional infrastructures or they are not suitable to be used in an indoor environment. Therefore, this study proposes using data from vision and inertial sensors which comprise 3-axis of accelerometer and 3-axis of gyroscope, also known as 6-degree of freedom (6-DOF) to determine pose estimation of mobile robot. The inertial measurement unit (IMU) based tracking provides fast response, therefore, they can be considered to assist vision whenever it fails due to loss of visual features. The use of vision sensor helps to overcome the characteristic limitation of the acoustic sensor for simultaneous multiple object tracking. With this merit, vision is capable of estimating pose with respect to the object of interest.

A singular sensor or system is not reliable to estimate the pose of a mobile robot due to limitations, therefore, data acquired from sensors and sources are combined using data fusion algorithm to estimate position and orientation within specific environment. The resulting model is more accurate because it balances the strengths of the different sensors. Information provided through sensor or data fusion can be used to support more-intelligent actions. The proposed algorithms are expedient to combine data from each of the sensor types to provide the most comprehensive and accurate environmental model possible. The algorithms use a set of mathematical equations that provides an efficient computational means to estimate the state of a process. This study investigates the state estimation methods to determine the state of a desired system that is continuously changing given some observations or measurements. From the performance and evaluation of the system, it can be observed that the integration of sources of information and sensors is necessary. This thesis has provided viable solutions to the challenging problem of localisation in autonomous mobile robot through its adaptability, accuracy, robustness and effectiveness.