Higher SVI → higher RSL → quicker roaming response. CCF ∈ [0,1] represents the penalty of premature roaming (e.g., re-authentication delay, data loss, monetary cost). For a video call, CCF is low (roaming is costly). For background sync, CCF is high (roaming is cheap). RSL is inversely related: ( RSL \propto (1 - CCF) ). 3.3 History-Dependent Hysteresis (HDH) HDH prevents ping-pong effects. Let ( R_past ) be the previous roaming time. Then:
where ( \alpha + \beta + \gamma = 1 ) (weighting coefficients determined by use case). SVI measures short-term fluctuations in primary link quality (e.g., RSSI, SNR). For ( N ) recent samples: roaming sensitivity level
[ RSL(t) = \alpha \cdot SVI(t) + \beta \cdot (1 - CCF(t)) + \gamma \cdot HDH(t) ] Higher SVI → higher RSL → quicker roaming response
We define as a dimensionless parameter, typically ranging from 0 (least sensitive, slowest to roam) to 1 (most sensitive, fastest to roam), that modulates the decision boundary for initiating a roaming event. RSL is not a single value but a dynamically adjustable state variable. 2. Related Work Existing mobility management protocols (e.g., MIH in IEEE 802.21, FMIPv6) use signal strength and latency thresholds but lack a unified sensitivity parameter. Reinforcement learning approaches adjust behavior post-facto, but none propose an explicit sensitivity level as a first-class control variable. Our work fills this gap by formalizing RSL and enabling predictive sensitivity tuning. 3. Mathematical Formulation of RSL Let the effective RSL at time ( t ) be defined as: For background sync, CCF is high (roaming is cheap)
Roaming Sensitivity Level, Handover Optimization, Context-Aware Computing, Mobility Management, Hysteresis Control. 1. Introduction Roaming—the process of transitioning a connection from one access point or service domain to another—is fundamental to mobile networks, IoT, and autonomous systems. Traditional roaming decisions rely on static thresholds (e.g., RSSI < -75 dBm triggers a scan). However, such rigidity fails in dynamic environments. Two identical signal drops may require opposite responses depending on user context, application sensitivity, or historical network reliability.