FSMP: A Frontier-Sampling-Mixed Planner for Fast Autonomous Exploration of Complex and Large 3-D Environments

TL;DR

FSMP addresses fast autonomous exploration of complex 3-D environments by integrating frontier-based and sampling-based strategies into a unified onboard planner. It introduces F^3D, a fast field-of-view–based frontier detector with completeness and soundness guarantees, and builds an incrementally updated road map via uniformly deterministic sampling. A two-stage planner performs a lazy, global optimal exploration path search on the road map and then applies path smoothing with time-optimal velocity profiling to improve efficiency. Across simulations and real-world experiments, FSMP achieves higher exploration speed, reduced computation time, and near-complete environmental coverage compared to state-of-the-art methods, demonstrating practical viability for large-scale MAV exploration.

Abstract

In this paper, we propose a systematic framework for fast exploration of complex and large 3-D environments using micro aerial vehicles (MAVs). The key insight is the organic integration of the frontier-based and sampling-based strategies that can achieve rapid global exploration of the environment. Specifically, a field-of-view-based (FOV) frontier detector with the guarantee of completeness and soundness is devised for identifying 3-D map frontiers. Different from random sampling-based methods, the deterministic sampling technique is employed to build and maintain an incremental road map based on the recorded sensor FOVs and newly detected frontiers. With the resulting road map, we propose a two-stage path planner. First, it quickly computes the global optimal exploration path on the road map using the lazy evaluation strategy. Then, the best exploration path is smoothed for further improving the exploration efficiency. We validate the proposed method both in simulation and real-world experiments. The comparative results demonstrate the promising performance of our planner in terms of exploration efficiency, computational time, and explored volume.

Figures (12)

• Figure 1: Illustration of the adjacency relationship of the map voxels. The red sphere indicates a specific voxel, and the green spheres denote its neighbors.
• Figure 2: Overall system architecture of the proposed exploration framework.
• Figure 3: The 2-D illustration of the detection process of F$^3$D. (a) The MAV is performing the exploration path, meanwhile, the FOVs of its onboard sensor are recorded. (b) There are overlapping areas between the FOVs. (c) F$^3$D can avoid repeatedly examining the voxels of the overlapping areas. (d) In this case, new frontiers are detected in the last recorded FOV.
• Figure 4: The 2-D schematic diagram of our uniformly deterministic sampling based incremental road map construction. (a) The exploration path can be found on the road map ${\cal R}$. (b) The MAV follows the exploration path while recording its FOVs. (c) All the FOVs that intersect with the same new frontier are extracted and merged into one AABB (i.e., the blue rectangles). In that region, uniform sampling is conducted. (d) The feasible new nodes (the yellow circles) and corresponding edges (the brown lines) are added to ${\cal R}$.
• Figure 5: The 2-D illustration of our two-stage exploration path planning method. (a) When the Dijkstra algorithm finds a new exploration candidate (the purple solid circle), the new radius of the Dijkstra's searching region (the region inside the red ring) will be computed. (b) The Dijkstra algorithm continues searching outward and finds the next new exploration candidate (the purple solid circle at the top right corner). Obviously, this new candidate has the maximum exploration gain, and thus no further searching is necessary. The reason is that the remaining candidates can not achieve a higher utility than this new candidate, since they are guaranteed to have a higher motion cost than this new candidate. (c) The global optimal exploration path is obtained, and we perform a path smoothing technique to facilitate the MAV's tracking.

Theorems & Definitions (6)

• Lemma 1
• proof
• Theorem 1
• proof
• Theorem 2
• proof