A Two-Stage Optimized Next-View Planning Framework for 3-D Unknown Environment Exploration, and Structural Reconstruction

Abstract

In this paper, we present a solution for autonomous exploration and reconstruction in 3-D unknown environments without a priori knowledge of the environments. In our framework, a two-stage heuristic information gain-based next-view planning algorithm is performed to dynamically select and update candidate viewpoints, followed by immediate next-best-view extraction and corresponding motion planning. The two-stage planner consists of a frontier-based boundary coverage planner and a fixed start open travelling salesman problem solver, such that the planner returns an exploration path with the consideration of global optimality in the context of accumulated space information. The effectiveness of our work is evaluated with simulation and experimental tests. The results and comparisons prove our system is able to autonomously explore the 3-D unknown environments and reconstruct the structural model with improved exploration efficiency in terms of path quality and total exploration time.