ARVC: A Closed-Loop, Resource-Aware Controller for Latency-Bounded IoT Dashboard Rendering on Constrained Edge Devices

Abstract

When the rendering parameters are fixed on a device with limited resources, frame drops and latency spikes are observed when the rendering parameters exceed its CPU capacity. We introduce a lightweight proportional–integral controller, embedded in the push/down sampling loop of a Flask/Socket, called ARVC (Adaptive Resource-Aware Visualization Controller).A real-time CPU-Memory-buffer backlog-round-trip latency IO dashboard to calculate a composite resource-pressure ψ(t) ∈. If ψ(t) is larger than a prescribed value (ψ_target = 0.55), ARVC discharges load by decreasing the number of points being rendered (LTTB down sampling), and by increasing the push interval; if the resources are nominal, ARVC restores full visual fidelity. We assessed ARVC on the Intel Berkeley Research Lab dataset replayed with the Raspberry Pi 4 (8 GB) at 50 ms intervals across the following 4 conditions: idle CPU, induced 4-core CPU-stress (23–25 independent trials of 30 s per condition; analysis was at the trial level to prevent pseudo replication). Under stress, ARVC raised frame‑delivery reliability by 0.99 percentage points (ACK rate 99.25% vs. 98.26%; 95% CI [+0.27, +1.68] pp; Holm–Bonferroni corrected p < 10⁻²¹), and it improved ACK rate by 1.28 pp under idle (95% CI [+0.66, +1.95] pp; p < 10⁻³⁴). The round-trip latency was statistically the same within 2ms (TOST p = 0.0155) at idle. No oscillation was seen during 3,768 verified control cycles on the controller. The results show that the closed-loop PI controller is able to reliably deliver the IoT dashboard frame on commodity edge hardware without changing protocols or hardware.