S5-SHB Agent: Society 5.0 enabled Multi-model Agentic Blockchain Framework for Smart Home

Janani Rangila; Akila Siriweera; Incheon Paik; Keitaro Naruse; Isuru Jayanada; Vishmika Devindi

cs.AIMar 5, 2026

S5-SHB Agent: Society 5.0 enabled Multi-model Agentic Blockchain Framework for Smart Home

Authors: Janani Rangila, Akila Siriweera, Incheon Paik, Keitaro Naruse, Isuru Jayanada, Vishmika Devindi

Organizations: KD University, Colombo, Sri Lanka · University of Aizu, Fukushima, Japan

Abstract

The smart home is a key application domain within the Society 5.0 vision for a human-centered society. As smart home ecosystems expand with heterogeneous IoT protocols, diverse devices, and evolving threats, autonomous systems must manage comfort, security, energy, and safety for residents. Such autonomous decision-making requires a trust anchor, making blockchain a preferred foundation for transparent and accountable smart home governance. However, realizing this vision requires blockchain-governed smart homes to simultaneously address adaptive consensus, intelligent multi-agent coordination, and resident-controlled governance aligned with the principles of Society 5.0. Existing frameworks rely solely on rigid smart contracts with fixed consensus protocols, employ at most a single AI model without multi-agent coordination, and offer no governance mechanism for residents to control automation behaviour. To address these limitations, this paper presents the Society 5.0-driven human-centered governance-enabled smart home blockchain agent (S5-SHB-Agent). The framework orchestrates ten specialized agents using interchangeable large language models to make decisions across the safety, security, comfort, energy, privacy, and health domains. An adaptive PoW blockchain adjusts the mining difficulty based on transaction volume and emergency conditions, using digital signatures and a Merkle tree to anchor transactions and ensure tamper-evident auditability. A four-tier governance model enables residents to control automation through tiered preferences from routine adjustments to immutable safety thresholds. Evaluation confirms that resident governance correctly separates adjustable comfort priorities from immutable safety thresholds across all tested configurations, while adaptive consensus commits emergency blocks.

Figures & tables

Figure 1: Gap-driven design motivation for S5-SHB-Agent

Symbol	Definition
$\delta_{t}$	Mining difficulty at block $t$
$\bar{v}_{t}$	Avg. transaction count over sliding window of $w$ blocks
$\mathcal{A}$	Set of specialized agents $\{a_{1},a_{2},\ldots,a_{n}\}$
$\mathcal{T}(t)$	Device telemetry vector at time $t$
$\pi(a_{i})$	Continuous priority score of agent $a_{i}$
$\sigma_{i}$	Role-specific system prompt for agent $a_{i}$

Table 1: Summary of notation.

References

Application Domain

Blockchain

AI / ML Integration

Human-Center Govern

Deployment Mode

SC/IoT

Fixed

Adapt

Cross

None

Deep

LLM

MultiA

MultiM

None

BAC

TierGov

S 5.0

Sim

Real

Hybrid

Multi

G1: [ 9 , 10 , 11 ]

\checkmark

–

\checkmark

–

[ 10 ]

–

[ 9 , 11 ]

–

[ 9 , 11 ]

[ 10 ]

–

\checkmark

–

G2: [ 12 ]

\checkmark

–

\checkmark

–

\checkmark

–

\checkmark

–

\checkmark

–

\checkmark

–

\checkmark

–

G3: [ 13 , 14 , 19 ]

–

\checkmark

–

\checkmark

–

[ 13 ]

[ 19 ]

[ 14 ]

–

\checkmark

–

\checkmark

\checkmark

–

G4: [ 16 , 17 , 15 , 18 ]

–

\checkmark

–

\checkmark

–

\checkmark

\checkmark

\checkmark

–

\checkmark

–

\checkmark

–

G5: [ 20 , 21 ]

–

\checkmark

–

\checkmark

–

\checkmark

–

[ 21 ]

[ 20 ]

–

\checkmark

–

Table 2: Grouped comparison of blockchain-based IoT framework capabilities.

Figure 2: System architecture of S5-SHB Agent showing the four-layer implementation with Control Plane, Agent Intelligence, Device & Data, and External layers.

Figure 3: Simulation View interface showing real-time threat scenario execution with home environment visualization and live event log streaming.

Agent $A_{i}$	Responsibility	Priority $\pi$	Output Domain
Safety Agent	Fire, gas, smoke response	Safety Critical (1.0)	Emergency commands
Arbitration Agent	Conflict resolution	Safety Critical (0.95)	Resolution decisions
Health Agent	Occupant wellness monitoring	High (0.9)	Health alerts
Anomaly Agent	ML-based anomaly detection	High (0.88)	Anomaly alerts
NLU Agent	Natural language parsing	High (0.85)	Structured commands
Security Agent	Intrusion detection	Medium (0.8)	Lock/alarm actions

Table 3: Agent Specification with Priority

Figure 4: Blockchain Explorer interface displaying the immutable transaction ledger with Ed25519-signed transactions, registered agents, and chain validity verification.

Figure 5: Agent Monitor interface showing multi-agent reasoning cycles with per-agent decisions, confidence scores, and real-time activity feed.

Tier	Scope	Parameters (Count)	Mutability
$G_{1}$ (Safe)	Routine adjustments	Temperature, brightness, quiet hours, voice, alerts (8 keys)	Resident-modifiable
$G_{2}$ (Impactful)	Trade-off settings	Comfort-vs-energy, security-vs-privacy, automation level (7 keys)	Resident-modifiable with validation
$G_{3}$ (Advanced)	System overrides	Per-agent device overrides, API budget caps, allowed providers (3 keys)	Resident-modifiable with validation
$G_{4}$ (Locked)	Safety invariants	Safety thresholds, crypto settings, override authority (9 keys)	Immutable

Table 4: Four-Tiered Governance Preference Hierarchy

Dim	Name	A Absent	$\triangle$ Partial	P Full
D1	Resident Authority	No user-configurable params	Binary permissions (grant/deny)	Graduated multi-tier control
D2	Safety Immutability	No safety protection	General blockchain tamper-resistance	Specific safety params locked at code level
D3	Governance Granularity	Single-level flat governance	2 levels (e.g. owner/user)	3+ graduated tiers with distinct permissions
D4	Audit Completeness	No audit trail	Some transaction logging	All governance mutation types audited
D5	Validation Coverage	No input validation	Identity/permission checks only	Type + bound validation on mutable params
D6	Conflict Resolution	No mechanism	Simple priority-based	Multi-level arbitration cascade

Table 5: Society 5.0 governance compliance dimensions and three-level rubric.

Dimension	Ours	[ 4 ]	[ 17 ]	[ 18 ]	[ 21 ]	[ 22 ]
D1 Resident Authority	$\checkmark$	$\triangle$	$\triangle$	$\triangle$	$\times$	$\times$
D2 Safety Immutability	$\checkmark$	$\triangle$	$\triangle$	$\triangle$	$\triangle$	$\checkmark$
D3 Governance Granularity	$\checkmark$	$\triangle$	$\triangle$	$\triangle$	$\triangle$	$\times$
D4 Audit Completeness	$\checkmark$	$\triangle$	$\triangle$	$\triangle$	$\triangle$	$\triangle$
D5 Validation Coverage	$\triangle$	$\triangle$	$\triangle$	$\triangle$	$\triangle$	$\triangle$
D6 Conflict Resolution	$\checkmark$	$\times$	$\times$	$\times$	$\times$	$\times$

Table 6: Governance Feature-Presence Comparison

Config	Type	Latency (ms/blk)	Throughput (tx/s)	Memory (KB)	Storage (KB)	Emergency (ms)
A	Static d=2	19.3	291.7	61.9	55.1	43.3
B	Static d=3	243.1	26.0	61.9	55.1	590.2
C	Adaptive (balanced)	477.8	15.9	61.7	55.1	15.2
D	Adaptive (aggressive)	4641.9	1.5	61.7	55.1	5.4
E	Adaptive (conservative)	636.9	12.2	61.8	55.1	191.7

Table 7: Our measured blockchain performance across five configurations.

Panel of Figure 6 — Figure 6: Agent priority contour surfaces over governance grid. Subplots show one adjustable agent’s priority as function: comfort_vs_energy (horizontal) & security_vs_privacy (vertical).

Panel of Figure 7 — Figure 7: Priority trajectories of all ten agents along each governance slider axis. Invariant agents (dashed) maintain fixed priorities; adjustable agents (solid) respond to slider changes.

Panel of Figure 8 — Figure 8: Per-block mining time distribution across five configurations (A-E) for each operational phase. Adaptive configurations reduce difficulty during EMERGENCY, yielding sub-15 ms block commits compared to 590 ms for static $d$ =3.

Panel of Figure 9 — Figure 9: Computing cost of mining across five configurations. (a) End-to-end mining time for a 108-txn run. (b) Total hash iterations required. (c) Resulting txn throughput. Config C (balanced) achieves a practical trade-off at 9.6 s and 16 tx/s, while Config D (aggressive) consumes 93 s for only 2 tx/s.

Panel of Figure 10 — Figure 10: Block-by-block difficulty traces across the four operational phases (IDLE, NORMAL, EMERGENCY, RECOVERY). Solid lines show median difficulty over 10 runs; shaded bands show IQR. Dashed lines mark static baselines A ( $d{=}2$ ) and B ( $d{=}3$ ).

Panel of Figure 11 — Figure 11: Multi-agent decision quality against Gemini model variants. (a) Box plots of per-decision confidence scores from baseline sessions (b) Number of decisions issued by each agent role, comparing baseline (no faults) against threat (7 injected faults), aggregated across all models.

Panel of Figure 12 — Figure 12: Blockchain behavior under multi-model agent load. (a) Average number of txns committed per block for each Gemini variant, (b) Difficulty level at each block index, shows Flash holds a stable $d{=}2$ while 2.0 Flash and Flash Lite oscillate between $d{=}1$ and $d{=}2$ in response to their higher txn rates.

Panel of Figure 13 — Figure 13: Blockchain mining integrity across 910 blocks (5 baseline + 5 threat sessions). (a) Nonce distribution with median of 196 iterations; baseline and threat overlap almost entirely. (b) Per-block nonce over block index showing no temporal drift in either condition.

Panel of Figure 14 — Figure 14: Agent decision reliability over 10 repeated sessions. (a) All sessions achieve 100% decision acceptance rate across 3,916 decisions. (b) Confidence distributions share the same median (0.90) under both conditions, confirming that threat injection does not degrade reasoning quality.

Panel of Figure 15 — Figure 15: End-to-end performance profile. (a) Safety-critical agents respond fastest (median 743 ms); maintenance is slowest (median 4,131 ms). (b) Anchor blocks hold exactly 1 transaction while decision blocks carry 5–17, with both bands stable across 90 blocks.

AI Agent Governance

AI agent governance studies how to authorize, constrain, monitor, and hold autonomous and multi-agent systems accountable.

See all 442 papers

Mar 2, 2026

S5-HES Agent: Society 5.0-driven Agentic Framework to Democratize Smart Home Environment Simulation

The smart home is a key domain within the Society 5.0 vision for a human-centered society. Smart home technologies rapidly evolve, and research should diversify while remaining aligned with Society 5.0 objectives. Democratizing smart home research would engage a broader community of innovators beyond traditional limited experts. This shift necessitates inclusive simulation frameworks that support research across diverse fields in industry and academia. However, existing smart home simulators require significant technical expertise, offer limited adaptability, and lack automated evolution, thereby failing to meet the holistic needs of Society 5.0. These constraints impede researchers from efficiently conducting simulations and experiments for security, energy, health, climate, and socio-economic research. To address these challenges, this paper presents the Society 5.0-driven Smart Home Environment Simulator Agent (S5-HES Agent), an agentic simulation framework that transforms traditional smart home simulation through autonomous AI orchestration. The framework coordinates specialized agents through interchangeable large language models (LLMs), enabling natural-language-driven end-to-end smart home simulation configuration without programming expertise. A retrieval-augmented generation (RAG) pipeline with semantic, keyword, and hybrid search retrieves smart home knowledge. Comprehensive evaluation on S5-HES Agent demonstrates that the RAG pipeline achieves near-optimal retrieval fidelity, simulated device behaviour and threat scenarios align with real-world IoT datasets, and simulation engine scales predictably across home configurations, establishing a stable foundation for Society 5.0 smart home research. Source code is available under the MIT License at https://github.com/AsiriweLab/S5-HES-Agent.

Akila Siriweera, Janani Rangila, Keitaro Naruse +2

The University of Aizu, Aizu Wakamatsu, Fukushima, Japan · The KD University, Rathmalana, Colombo, Sri Lanka

Internet of Things LLM Agent Orchestration

Ask PDF

Mar 16, 2026

HearthNet: Edge Multi-Agent Orchestration for Smart Homes

Smart-home users increasingly want to control their homes in natural language rather than assemble rules, dashboards, and API integrations by hand. At the same time, real deployments are brittle: devices fail, integrations break, and recoveries often require manual intervention. Existing agent toolkits are effective for session-scoped delegation, but smart-home control operates under a different scenario: it is persistent, event-driven, failure-prone, and tied to physical devices with no shared context window. We present HearthNet, an edge multi-agent orchestration system for smart homes. HearthNet deploys a small set of persistent, role-specialized LLM agents at the home hub, where they coordinate through MQTT, Git-backed shared state, and root-issued actuation leases to govern heterogeneous devices through thin adapters. This design externalizes context, preserves execution history, and separates planning, verification, authorization, and actuation across explicit boundaries. Our current prototype runs on commodity edge hardware and Android devices; it keeps orchestration, state management, and device control on-premise while using hosted LLM APIs for inference. We demonstrate the system through three live scenarios: intent-driven multi-agent coordination from ambiguous natural language, conflict resolution with timeline-based tracing, and rejection of stale or unauthorized commands before device actuation.

Zhonghao Zhan, Krinos Li, Yefan Zhang +1

Imperial College London, London, UK · Independent Researcher, Seattle, USA

LLM Agent Orchestration Internet of Things Tool Access Control for LLM Agents

Ask PDF

May 31, 2026

HomeFlow: A Data Flywheel for Smart Home Agent Training with Verifiable Simulation

Large language model agents are moving beyond text-only interaction toward physical-world control, with smart homes as a representative domain. Real domestic interaction requires understanding ambiguous intents, operating in dynamic environments, and performing multi-turn reasoning. However, existing methods struggle to generate high-quality training data for smart home agents. We propose HomeFlow, a verifiable data flywheel for this domain. HomeFlow uses HomeEnv as a unified simulation environment and HomeMaker to procedurally generate diverse home settings. Subsequently, Blueprint compiles open-ended user intents into executable state-based success conditions, while MCTS-Flow synthesizes diverse, verifiable multi-turn trajectories through environment-guided tree search. We then optimize the agents via supervised fine-tuning and step-wise RLVE, which facilitates iterative improvement through authentic physical feedback. We further construct SmartHome-Bench to evaluate the agent across various smart home tasks. On this benchmark, HomeFlow-RL-4B and HomeFlow-RL-8B achieve task success rates of 84.60% and 87.03%. It is worth noting that HomeFlow-RL-8B even surpasses the leading GPT-5.5 by 1.23 percentage points.

Yi Gu, Huacan Wang, Shuo Zhang +10

1Midea Group · 2Beijing University of Posts and Telecommunications · 3Donghua University +1

AI Agent Benchmarks LLM Agent Training RL for LLM Agents

Ask PDF

Abstract

AI Agent Governance

Explore similar work

S5-HES Agent: Society 5.0-driven Agentic Framework to Democratize Smart Home Environment Simulation

HearthNet: Edge Multi-Agent Orchestration for Smart Homes

HomeFlow: A Data Flywheel for Smart Home Agent Training with Verifiable Simulation