Introduction
As Internet of Things (IoT) ecosystems expand from simple consumer gadgets to critical infrastructure like Intelligent HealthCare (IHC) systems, the complexity of resource management has scaled exponentially. Efficiently allocating resources—such as heart rate sensors, smart treadmills, and monitoring devices—to various healthcare services is a computationally "NP-Hard" problem. As the number of devices and service requests grows, traditional optimization methods often fail to find viable solutions in a reasonable timeframe.
This paper introduces BRAD-GWA, a novel framework designed to address two primary gaps in current IoT research: the lack of multi-scenario adaptability and the over-reliance on single-metric optimization (focusing solely on either profit or cost). By combining a hardware-agnostic abstraction layer with a biologically-inspired evolutionary algorithm, the authors provide a robust solution for diverse healthcare environments.
The BRAD Model: Abstraction and Bimetric Optimization
The proposed BRAD (Bimetric-balanced Resource Allocation with Digital objects) model rests on two pillars: hardware abstraction and balanced objectives.
Digital Objects (DO)
To manage the heterogeneity of IoT hardware, the system abstracts physical devices into "Digital Objects." These objects provide standardized APIs, allowing the allocation logic to remain hardware-agnostic. This ensures that the resource allocator interacts with a unified "Collaboration Module" rather than managing the specific technical requirements of individual sensors.
Bimetric Objective Function
Unlike traditional models that optimize for a single variable, BRAD utilizes a bimetric approach. The optimization goal is defined by the objective function:
$$Obj = Cost + w \times Profit$$
By adjusting the weight ($w$), system providers can fine-tune the balance between operational expenditure and revenue generation. This flexibility is crucial for multi-scenario deployments, such as balancing the needs of a resource-rich city hospital against a resource-scarce community clinic.
BRAD-GWA: Advancing the Grey Wolf Optimizer
To solve the complex allocation problem, the authors developed BRAD-GWA, an enhanced version of the Grey Wolf Optimizer (GWO). Standard GWO mimics the social hierarchy and hunting behavior of wolves (Alpha, Beta, and Delta) to find optimal solutions. However, standard GWO often suffers from premature convergence and a lack of population diversity. BRAD-GWA introduces three key biological improvements:
- Density-Aware Initialiser: To prevent the algorithm from starting with a cluster of near-identical solutions, this initialiser uses L2 distance metrics to merge similar search agents. This ensures the "wolves" are distributed broadly across the search space from the start.
- Enhanced Greedy Social Hierarchy: While standard GWO relies on the top three wolves, BRAD-GWA utilizes the top seven. It employs four distinct hierarchy structures to lead the pack, significantly reducing the risk of the algorithm getting trapped in "local optima" (sub-optimal solutions that appear favorable locally).
- Lifetime-Enabled Elimination: Mimicking natural selection, the algorithm assigns a "lifetime" to each agent. Poorly performing wolves that fail to improve are eliminated and replaced with new, randomized agents, maintaining the "freshness" of the search and encouraging exploration of new solution areas.
Experimental Evaluation
The framework was tested using the HIT-IHC (Intelligent HealthCare system) testbed developed by the Harbin Institute of Technology. The simulation included:
- Diverse Scenarios: A resource-rich city hospital and a resource-scarce community care home.
- Resources: 10 types of IoT devices with varying availability (3 to 50 units each) and associated costs.
- Services: 200 healthcare services (e.g., rehabilitation, continuous monitoring) simulated via the
Simpydiscrete event simulator.
Key Results
BRAD-GWA was benchmarked against 13 other methods, including Particle Swarm Optimization (PSO), the Whale Optimization Algorithm (WOA), and the original HIT-IHC logic. The results demonstrate significant improvements across all metrics:
- Optimization Performance: BRAD-GWA achieved a 3.14x reduction in the objective function compared to the original HIT-IHC method and a 29.6% reduction compared to the standard GWO.
- Economic Efficiency: The algorithm delivered 12.3% higher profit than standard GWO and achieved the highest Profit/Cost ratio (0.38), indicating superior resource efficiency.
- Service Delivery: It successfully allocated an average of 81.6 services, the highest throughput among all tested algorithms.
Conclusion and Future Work
The BRAD-GWA framework demonstrates that maintaining population diversity and employing scenario-aware bimetric optimization can significantly enhance IoT resource allocation. By abstracting hardware through Digital Objects and refining evolutionary search mechanics, the system achieves a more efficient and profitable balance than existing metaheuristics.
Future research will focus on expanding the model beyond binary scenarios (rich vs. scarce) to a broader spectrum of resource levels. Additionally, the authors plan to investigate dynamic re-allocation to handle real-time "streaming" requests, ensuring the system can adapt as patients move or emergencies arise in a live healthcare environment.
Summary updated