CHAOS - Configurations Analysis of Swarms of Cyber-Physical Systems

Valeria Trombetta, Maxime Cordy, Enrico Tronci, Axel LegayCPS Summer School, PhD Workshop 2022 · OpenReview

Introduction

As Industry 4.0 matures, Cyber-Physical Systems (CPS)—integrated networks of computational and physical components—are becoming ubiquitous. These systems are often "variability-intensive," meaning they can be configured in millions of ways through different hardware components and software versions. However, analyzing these configurations is difficult due to the massive state space and the unpredictable nature of real-world environments. The CHAOS framework (Configurations Analysis of Swarms of Cyber-Physical Systems) was developed to bridge this gap.

The Challenge: State Space and Uncertainty

The CHAOS framework addresses two primary hurdles in CPS design:

  1. State Space Explosion: Even a small swarm of drones with a few configurable features creates a configuration space too large for exhaustive manual analysis.
  2. The Uncertainty Gap: Traditional configuration tools often assume ideal conditions. In reality, CPS must navigate "noise" (sensor errors), "gusts" (environmental disturbances), and "faults" (hardware failures). A configuration optimized for a vacuum may fail in the field.

Methodology: The CHAOS Framework

The authors developed a simulation-based framework using MATLAB and Simulink to model a swarm of up to nine drones. The mission involves navigating an obstacle-filled arena to reach a target.

Variability and Uncertainty Modeling

The study modeled variability by assigning drones different quality levels (Top, Medium, Low) for two key features: Battery and Radio. This allowed for heterogeneous swarms with mixed component tiers. To bridge the uncertainty gap, the framework injected three layers of stochastic disturbances:

  • White Noise: Simulating mechanical jitter in thrusters.
  • Random Gusts: Simulating wind impact on trajectories.
  • Markov Chain Faults: Simulating intermittent radio communication failures.

The framework utilizes Monte Carlo simulations (50 runs per configuration) to verify if a setup meets a 96% "assurance level" for mission success across a total of 12,264 unique scenarios.

Key Insights and Results

The analysis yielded several critical findings regarding how environmental stress impacts system design:

  • Environmental Sensitivity: While white noise and radio faults were found to be manageable, wind gusts were transformative. Gusty environments almost always necessitated higher-quality components to maintain the 96% success rate.
  • The Cost of Mobility: If the mission target is moving, the "cost" of a successful configuration increases significantly (by an average of 7 units), often requiring an increase in the number of drones rather than just component upgrades.
  • Optimization over Overkill: The study found that the "best" configuration is rarely just the most expensive one. By analyzing cross-configurations, CHAOS identified ways to save budget by using medium-tier components while still hitting the required success target.

Conclusion and Future Work

The CHAOS framework demonstrates that variability monitoring is essential even for relatively small systems. By quantifying the impact of environmental stress, designers can avoid "overkill" and select configurations that are both robust and cost-effective.

While the current approach is computationally intensive, future work focuses on a "snapshot" method to stop simulations early if a configuration mirrors previously tested data. Additionally, the authors plan to expand the framework to more complex mission types and other CPS domains, such as miniature satellite (CubeSat) constellations.

Summary updated

BibTeX
@inproceedings{trombetta2022chaos,
  author = {Valeria Trombetta and Maxime Cordy and Enrico Tronci and Axel Legay},
  title = {{CHAOS - Configurations Analysis of Swarms of Cyber-Physical Systems}},
  booktitle = {CPS Summer School, PhD Workshop},
  year = {2022},
  url = {https://openreview.net/forum?id=sxerJ0oQgV},
}
Figure