Adaptive Polling Mechanisms for Real-Time Polling Systems: From Cyclic to Arrival-Rate-Based Approaches

Polling systems are critical components in various queue-based communication environments such as wireless sensor networks, industrial automation, and embedded systems. The choice of an efficient polling strategy significantly affects system performance in terms of latency, throughput, and resource utilization. This paper provides a comprehensive comparative analysis of three representative polling strategies: traditional cyclic polling, adaptive skip-empty polling, and a novel adaptive polling algorithm based on real-time arrival rate estimation. The proposed approach leverages lightweight, on-the-fly traffic measurements to prioritize queues with higher estimated arrival rates, dynamically adjusting the polling schedule without requiring prior statistical knowledge or predictive modeling. Unlike static or threshold-based methods, our algorithm is inherently responsive to fluctuations in traffic patterns and avoids unnecessary polling overhead. We implement a simulation framework to evaluate the performance of all strategies under various traffic conditions, including bursty and unbalanced loads. The results reveal that the arrival-rate-based strategy consistently outperforms the alternatives in reducing average queueing delay and enhancing load balancing. Additionally, we examine the trade-offs between responsiveness and computational overhead, demonstrating the suitability of our method for real-time and resource-constrained systems. The findings suggest that adaptive polling guided by arrival rate estimation offers a promising direction for designing scalable and delay-sensitive scheduling algorithms.