PPO-HAHH: a proximal policy optimization-based hierarchical adaptive hyper-heuristic for mobile robot path planning
Year: 2026
Type: article
Abstract: Mobile robot path planning in complex grid environments is often hampered by obstacle constraints and fragmented feasible regions, leading to local optima and unstable convergence. To enhance adaptive search scheduling under complex constraints, this paper proposes a Proximal Policy Optimization (PPO)-based Hierarchical Adaptive Hyper-Heuristic algorithm, termed PPO-HAHH. The algorithm adopts a two-level framework consisting of a high-level policy and low-level heuristics. At the high level, PPO produces a one-dimensional continuous scheduling signal from population statistical features and maps it to dynamic weights of exploration-oriented, balance-oriented, and exploitation-oriented operator pools to balance global exploration and local exploitation. At the low level, ten complementary heuristic operators are constructed, and an Individual-level Strategy Preference (ISP) mechanism is introduced to jointly determine operator selection probabilities with operator-pool weights, enabling individualized search under unified macro scheduling. Experiments on the CEC2017 benchmark suite and fixed and random grid maps show that PPO-HAHH is competitive with the selected baselines in terms of overall optimization rank, path length, effective turn count, and feasible-path success rate. Analysis of policy behavior further shows that the scheduler adjusts operator-pool weights with search progress, map scale, and obstacle coverage.
Authors: Yadong Bai, Qingni Yuan, Wei Zhang, Pengju Qu
Institution: Guizhou University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $1,990
Award: G3776940148
OpenAlex ID: W7225751517