Ontology-Supported Global Path Planning and Real-Time Navigation for a Humanoid Robot

Path planning is a fundamental requirement for safe robot navigation and task execution in complex environments. This paper proposes an ontology-supported path planning and real-time navigation approach for the NAO humanoid robot. The proposed architecture separates global path planning from real-time path execution while integrating both stages through a common ontology-based representation of the robot and its environment. An Enhanced A* planner, incorporating an occupancy-grid representation, obstacle safety regions, and safe approach-goal selection, is employed once, prior to robot motion, to generate a global waypoint path toward the target object. The resulting path is subsequently executed by a custom robot controller, while the ontology maintains and updates information about the robot and the objects in the workspace throughout navigation. Path execution is complemented by reactive sonar-based obstacle avoidance for handling obstacles without invoking global replanning, and direct movement toward the target when there is a free path. Experimental evaluation in both simulated and real-world environments using the NAO humanoid robot demonstrates the feasibility and effectiveness of the proposed approach. More specifically, comparison of our enhanced A* with A*, Dijkstra and Inflated grid A* in the simulated experiments showed its clear superiority in task success (83.43% vs. 36.71% of the next best) with very comparable average execution time (0.997 ms vs. 0.707 ms of the next best). Also, a 100% task success rate was recorded in our real-time experiments with the NAO robot in our 180 × 240 cm workspace for configurations with up to four obstacles.