TY - GEN
T1 - EC-HSRA
T2 - 8th International Conference on Electronic Engineering and Informatics, EEI 2026
AU - Ma, Jingxiang
AU - Khan, Malak Abid Ali
AU - Ma, Hongbin
AU - Zhang, Youzhi
N1 - Publisher Copyright:
© 2026 IEEE.
PY - 2026
Y1 - 2026
N2 - Regional support and emergency scheduling require repeated allocation of scarce defense, repair, mobility-assurance, and reconnaissance resources across responsibility regions under changing conditions. A shared-budget decision may prioritize a relevant task while failing to cover task-specific regions, or allocate resources to salient regions under an inappropriate task. Fixed sharing, pressure gating, and direct high-dimensional policies can ensure budget feasibility, yet they do not explicitly couple task allocation with task-conditioned regional decoding. This paper proposes Environment-Conditioned Hierarchical Shared-Resource Allocation (EC-HSRA). EC-HSRA abstracts the operating area as a regional graph and uses a constrained hierarchical action: the task layer learns state-conditioned task-budget allocation, and the regional layer generates task-specific threshold, focus, and coverage-mixing parameters. In the situation-cascade scenario, EC-HSRA reaches a mean shifted cumulative utility of 17.425, exceeding Pressure Gate at 10.972, Top-k Stitch at 14.522, and the explicit-prior variant at 16.524. Its normalized service-shortfall index is 0.941, lower than Pressure Gate at 1.047 and Unified Joint at 1.147. The results indicate that task-level allocation and task-specific regional decoding are complementary.
AB - Regional support and emergency scheduling require repeated allocation of scarce defense, repair, mobility-assurance, and reconnaissance resources across responsibility regions under changing conditions. A shared-budget decision may prioritize a relevant task while failing to cover task-specific regions, or allocate resources to salient regions under an inappropriate task. Fixed sharing, pressure gating, and direct high-dimensional policies can ensure budget feasibility, yet they do not explicitly couple task allocation with task-conditioned regional decoding. This paper proposes Environment-Conditioned Hierarchical Shared-Resource Allocation (EC-HSRA). EC-HSRA abstracts the operating area as a regional graph and uses a constrained hierarchical action: the task layer learns state-conditioned task-budget allocation, and the regional layer generates task-specific threshold, focus, and coverage-mixing parameters. In the situation-cascade scenario, EC-HSRA reaches a mean shifted cumulative utility of 17.425, exceeding Pressure Gate at 10.972, Top-k Stitch at 14.522, and the explicit-prior variant at 16.524. Its normalized service-shortfall index is 0.941, lower than Pressure Gate at 1.047 and Unified Joint at 1.147. The results indicate that task-level allocation and task-specific regional decoding are complementary.
KW - hierarchical reinforcement learning
KW - multi-task regional support
KW - regional allocation
KW - shared-resource allocation
KW - task-region mismatch
UR - https://www.scopus.com/pages/publications/105048971165
U2 - 10.1109/EEI70303.2026.11640621
DO - 10.1109/EEI70303.2026.11640621
M3 - Conference contribution
AN - SCOPUS:105048971165
T3 - 2026 8th International Conference on Electronic Engineering and Informatics, EEI 2026
SP - 172
EP - 176
BT - 2026 8th International Conference on Electronic Engineering and Informatics, EEI 2026
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 26 June 2026 through 28 June 2026
ER -