Adaptive Stream Processing on Edge Devices through Active Inference

TL;DR

The paper tackles the challenge of maintaining QoS for real-time stream processing on resource-constrained edge devices under dynamic conditions. It introduces an Adaptive Inference (AIF) agent that builds and continually updates a causal generative model to predict SLO fulfillment and to select local configurations that minimize Expected Free Energy, balancing pragmatic performance and knowledge gain. The approach emphasizes transparency through causal structures and homeostatic principles, and it validates the method on three edge-based streaming services across heterogeneous edge hardware, showing convergence in up to about 30 iterations while preserving interpretability. The work demonstrates the practical viability of decentralized, AIF-driven elasticity for edge streaming, with potential for broader applicability and future comparisons to alternative ML-based adaptation strategies.

Abstract

The current scenario of IoT is witnessing a constant increase on the volume of data, which is generated in constant stream, calling for novel architectural and logical solutions for processing it. Moving the data handling towards the edge of the computing spectrum guarantees better distribution of load and, in principle, lower latency and better privacy. However, managing such a structure is complex, especially when requirements, also referred to Service Level Objectives (SLOs), specified by applications' owners and infrastructure managers need to be ensured. Despite the rich number of proposals of Machine Learning (ML) based management solutions, researchers and practitioners yet struggle to guarantee long-term prediction and control, and accurate troubleshooting. Therefore, we present a novel ML paradigm based on Active Inference (AIF) -- a concept from neuroscience that describes how the brain constantly predicts and evaluates sensory information to decrease long-term surprise. We implement it and evaluate it in a heterogeneous real stream processing use case, where an AIF-based agent continuously optimizes the fulfillment of three SLOs for three autonomous driving services running on multiple devices. The agent used causal knowledge to gradually develop an understanding of how its actions are related to requirements fulfillment, and which configurations to favor. Through this approach, our agent requires up to thirty iterations to converge to the optimal solution, showing the capability of offering accurate results in a short amount of time. Furthermore, thanks to AIF and its causal structures, our method guarantees full transparency on the decision making, making the interpretation of the results and the troubleshooting effortless.

Figures (8)

• Figure 1: Abstract representation of a continuous stream processing scenario; sensors provide input data that is processed by services located at an edge device; processing is observable through metrics and respective results are provided to stream consumers
• Figure 2: Conditional (or causal) variable relations encoded in a Bayesian network; variable states (i.e., fps and pixel) form a 2D solution space where each parameter combination features a distinct pragmatic value (pv) and information gain (ig)
• Figure 3: High-level action-perception cycle in AIF; stream processing metrics are interpreted using a generative model; the agent updates the model according to prediction errors and makes adjustments by balancing pragmatic value with information gain
• Figure 4: Demo output for each service according to prerecorded input data; all three services are iteratively processing video streams (CV and QR) or binary input (LI)
• Figure 5: Empirical SLO fulfillment measured during service execution; initial training rounds show unstable behavior, while later rounds converge to a clear preference