RRT Based Frontier Point Detection for 2D Autonomous Exploration

Abstract

The ability to explore an unknown environment and create a representation is a must-have for a fully autonomous robot. Autonomous exploration is a multifaceted problem and requires solutions to a combination of sub-problems such as Simultaneous Localization and Mapping (SLAM), path planning and following, detecting potential navigation target points and evaluating them to select a target. This study implements a frontier point based approach to identify potential navigation targets. Permanent and temporary RRT-based tree structures are used to search the existing map and detect frontier points. Permanent tree is rooted from the initial position of the robot and is never reset. Temporary trees are reset when they hit a frontier point or reach a certain number of branches. Two types of temporary trees are used, starting from the robot's current location and starting from the frontier points that lose their frontier point characteristics as the region they lie is mapped. A revenue function considering path cost, heading cost, and information gain is used to evaluate the frontier points and select a target among them.