A Comparison into Current SLAM Implementations

Automation has recently had a new increase in interest given its applications with the surge in machine learning innovations, self-driving cars, and more accessible computing capabilities. An integral part of automation is understanding an environment to operate within it. To understand environments we rely upon maps to traverse through them. In most real-world applications these environments are dynamic and ever-changing, thus a static map cannot be relied upon to completely describe an environment. SLAM attempts to solve this through continuous mapping and simultaneous localization within an environment. Finding applications in many fields SLAM still has many challenges to overcome with each implementation having its individual strengths. SLAM implementations are explored both in literature and with real-world testing as an initial study into what SLAM entails and the methods currently existing. Turtlebot 3 is a robotics platform used to explore this field and is utilized to create a map of a known lab environment. Overall newer implementations come with improvements in accessibility, costs, and accuracy, but still come with flaws including low resolution depending on sensors, and increased computational costs.