2022 IB Diploma Extended Essays

determine it’s location within an environment and plan a path that will take the system to it’s goal [1]. Autonomous systems achieve navigation through the implementation of Simultaneous Localisation and Mapping (SLAM) SLAM SLAM is the process that allows a mobile robot to be able to localise itself and reconstruct an unknown environment using on board sensors [2]. The task is solved through the use of different sensors. The choice of which sensor is to implemented depends hugely on the task that needs to be solved. The choice of suitable sensors for autonomous ground vehicles (AGV) plays a important role in it’s efficiency due to it’s limited autonomy resource [3]. Most mobile robots contain a inertial measurement unit (IMU), which has a gyroscope, accelerometer and a magnetometer. Using a IMU makes it possible to measure angular velocity, orientation and acceleration [3]. This approach to autonomy by measuring the position of the mobile robot is known as Inertial Navigation System [3]. However, the aren’t solely relied upon to achieve autonomy due to it’s a large range of error. Visual SLAM Another widely used sensor in SLAM are camera’s. This is because camera’s are a good sensor due to small cost, simplicity in configuration process and higher quality reconstruction [4]. Solving SLAM through solely relying on the camera as the source of information about the surrounding scene is referred to as Visual SLAM [5]. The two most common methods of Visual SLAM are: Feature-based, which use the selected features from a scene to build maps and Direct-dense based SLAM, which reconstructs a map of the scene from a whole image. The first work on Visual SLAM involved a binocular stereo camera [6][7]. Similarly the problem of Visual SLAM was also implemented through a monocular camera, this method was called Mono-SLAM [8]. The investigation explores ORB SLAM which is a feature-based visual SLAM algorithm [13]. The main critique include weak resistance to bad conditions, poor performance with weak geometric diversity of environment, experimentation and sensitivity to poor rotations [22][23].