ContextWeave: Adaptive Context-Aware Lifecycle Orchestration for Privacy-Preserving Ephemeral Edge Computing

Edge orchestration research optimizes where tasks and services run, but usually assumes that the edge infrastructure itself is always present. Under mobile, bursty and event-driven demand, this assumption wastes energy on idle capacity and leaves sensitive service state resident long after the demand that justified it has ended. We treat the existence of the edge execution substrate as an orchestration variable and present CONTEXT-WEAVE, a context-aware lifecycle controller for ephemeral edge sessions. CONTEXTWEAVE observes only aggregate, privacy-minimized signals (arrival rates, trends, sensitive-task share and aggregate mobility inflow), forecasts short-horizon demand with an interpretable rule layer and an optional federated residual learner, converts the forecast and its uncertainty into the probability that a session has positive net value, and creates, extends or dissolves sessions with asymmetric evidence, hysteresis, minimum retention and cooldown.