Closed-Polygon Free Space Representation and Subgoal Switching Regulation for robust mapless robot navigation

Zhang, Ziqin, You, Yingchao, Jian, Shaojie, Wei, Changyun and Ji, Ze ORCID: https://orcid.org/0000-0002-8968-9902 2026. Closed-Polygon Free Space Representation and Subgoal Switching Regulation for robust mapless robot navigation. Robotics and Autonomous Systems 206 , 105751. 10.1016/j.robot.2026.105751

Full text not available from this repository.

Abstract

Reliable navigation in cluttered indoor environments remains challenging for mobile robots operating under partial observability. Static structures can occlude parts of the environment, while dynamic obstacles may appear unpredictably in the robot’s sensing range, together leading to an incomplete perception of free space. In reactive planners, this perceptual uncertainty often causes unstable behaviors such as oscillatory steering, repeated hesitation, or deadlock. This paper presents a lightweight, fully onboard navigation framework that addresses these issues through two coupled mechanisms: Closed-Polygon Free Space Representation and Subgoal Switching Regulation. First, we construct a closed, safety-inflated local traversable polygon from raw LiDAR scans. The polygon distinguishes hit-supported, free-confirmed, and unknown-closure boundary segments. Second, we introduce a Subgoal Switching Regulation Mechanism that improves temporal consistency in subgoal selection. The mechanism groups candidate subgoals into directional hypotheses and regulates switching between them with short-term statistics, thereby reducing unnecessary subgoal switching under changing observations. Simulation experiments show that the complete proposed pipeline achieves the highest or joint-highest success rates and the lowest command-level behavioral instability across the tested scenarios among the evaluated methods. Five repeated real-world deployments further demonstrate onboard feasibility and task-level repeatability in the tested indoor setting without requiring a prior global map.

Item Type:	Article
Date Type:	Publication
Status:	Published
Schools:	Schools > Engineering
Publisher:	Elsevier
ISSN:	0921-8890
Date of Acceptance:	10 September 2026
Last Modified:	05 Oct 2026 10:00
URI:	https://orca.cardiff.ac.uk/id/eprint/189978

Actions (repository staff only)

Edit Item