[Paper] Analysis of Asset Administration Shell-based Negotiation Processes for Scaling Applications

Source: arXiv - 2602.13029v1

Overview

The paper Analysis of Asset Administration Shell‑based Negotiation Processes for Scaling Applications examines how the proactive Asset Administration Shell (AAS) – a cornerstone of Industry 4.0 – handles negotiation traffic when the number of connected assets grows from a handful to thousands. By measuring message load, latency, and adaptability, the authors reveal where the current AAS specifications start to strain under realistic industrial loads.

Key Contributions

• Scalable evaluation framework – a reproducible scenario and a set of quantitative criteria (message count, round‑trip time, CPU/memory usage) for testing proactive AAS negotiations.
• Reference implementation – a fully functional prototype built on existing AAS architectures (submodel‑based data model, VDI/VDE 2193 language) that can be deployed with any number of virtual assets.
• Empirical performance study – systematic experiments ranging from 10 to 10 000 assets, exposing the communication overhead and bottlenecks of the negotiation protocol.
• Guidelines for standardization – concrete recommendations on message batching, subscription management, and security extensions to improve scalability in future AAS releases.

Methodology

1. Scenario design – The authors model a typical production‑planning use case: a “product requestor” negotiates with multiple “resource providers” (machines, robots, storage units) to allocate production slots.
2. Implementation stack –
   • AAS core: Eclipse AAS Toolkit (Java) with the VDI/VDE 2193 negotiation language.
   • Transport: MQTT 5.0 for bidirectional messaging, TLS for security.
   • Orchestration: Docker‑Compose to spin up thousands of lightweight AAS containers representing individual assets.
3. Metrics collection – Custom probes record the number of negotiation messages, average latency per negotiation round, and host resource consumption (CPU, RAM, network I/O).
4. Scaling runs – Experiments are executed with 10, 100, 1 000, 5 000, and 10 000 simulated assets, each repeated three times to ensure statistical confidence.

Results & Findings

• Message explosion: As assets increase, the number of negotiation messages grows non‑linearly because each requestor must broadcast to all potential providers and handle individual acknowledgments.
• Latency bottleneck: Beyond ~1 000 assets, round‑trip times exceed 100 ms, which can jeopardize real‑time scheduling in fast production lines.
• Resource saturation: CPU and network usage become the limiting factors; the MQTT broker’s internal queue starts to overflow at ~5 000 concurrent connections.
• Adaptability: Introducing a simple “interest‑filter” (providers only subscribe to relevant product categories) cuts messages by ~30 % and restores latency to sub‑100 ms even at 10 000 assets.

Practical Implications

• System architects can use the paper’s benchmark suite to stress‑test their own AAS deployments before rolling out to large factories.
• Middleware vendors should consider built‑in support for selective subscription, message aggregation, and back‑pressure mechanisms to keep MQTT brokers from choking under massive asset fleets.
• Developers building proactive AAS services can adopt the recommended “interest‑filter” pattern or hierarchical negotiation (edge‑gateway aggregators) to keep communication overhead manageable.
• Standard bodies (e.g., Plattform Industrie 4.0) gain concrete data to justify extensions to VDI/VDE 2193, such as optional batch negotiation primitives or lightweight “heartbeat” messages that reduce chatter.

Limitations & Future Work

• The experiments rely on a simulated environment (Docker containers) and a single MQTT broker; real‑world deployments may involve multiple brokers, different transport layers (OPC UA, AMQP), or edge‑gateway hierarchies.
• Security overhead (TLS handshake, token validation) was kept minimal; a full‑scale security analysis could reveal additional latency spikes.
• Future research suggested by the authors includes: (1) testing hierarchical negotiation topologies, (2) exploring alternative messaging protocols (e.g., DDS), and (3) integrating AI‑driven negotiation strategies that dynamically prune the search space of providers.

Authors

• David Dietrich
• Armin Lechler
• Alexander Verl

Paper Information

• arXiv ID: 2602.13029v1
• Categories: cs.SE
• Published: February 13, 2026
• PDF: Download PDF