Browsing by Author "Zou, Mengbang"

Now showing 1 - 9 of 9
  • Thumbnail Image
    ItemOpen Access
    Analysing region of attraction of load balancing on complex network
    (Oxford Academic, 2022-06-29) Zou, Mengbang; Guo, Weisi
    Many complex engineering systems network together functional elements to balance demand spikes but suffer from stability issues due to cascades. The research challenge is to prove the stability conditions for any arbitrarily large and dynamic network topology with any complex balancing function. Most current analyses linearize the system around fixed equilibrium solutions. This approach is insufficient for dynamic networks with multiple equilibria, for example, with different initial conditions or perturbations. Region of attraction (ROA) estimation is needed in order to ensure that the desirable equilibria are reached. This is challenging because a networked system of non-linear dynamics requires compression to obtain a tractable ROA analysis. Here, we employ master stability-inspired method to reveal that the extreme eigenvalues of the Laplacian are explicitly linked to the ROA. This novel relationship between the ROA and the largest eigenvalue in turn provides a pathway to augmenting the network structure to improve stability. We demonstrate using a case study on how the network with multiple equilibria can be optimized to ensure stability.
  • Thumbnail Image
    ItemOpen Access
    Explainable data-driven Q-learning control for a class of discrete-time linear autonomous systems
    (Elsevier, 2024-11-01) Perrusquía, Adolfo; Zou, Mengbang; Guo, Weisi
    Explaining what a reinforcement learning (RL) control agent learns play a crucial role in the safety critical control domain. Most of the approaches in the state-of-the-art focused on imitation learning methods that uncover the hidden reward function of a given control policy. However, these approaches do not uncover what the RL agent learns effectively from the agent-environment interaction. The policy learned by the RL agent depends in how good the state transition mapping is inferred from the data. When the state transition mapping is wrongly inferred implies that the RL agent is not learning properly. This can compromise the safety of the surrounding environment and the agent itself. In this paper, we aim to uncover the elements learned by data-driven RL control agents in a special class of discrete-time linear autonomous systems. Here, the approach aims to add a new explainable dimension to data-driven control approaches to increase their trust and safe deployment. We focus on the classical data-driven Q-learning algorithm and propose an explainable Q-learning (XQL) algorithm that can be further expanded to other data-driven RL control agents. Simulation experiments are conducted to observe the effectiveness of the proposed approach under different scenarios using several discrete-time models of autonomous platforms.
  • Thumbnail Image
    ItemOpen Access
    Explaining data-driven control in autonomous systems: a reinforcement learning case study
    (IEEE, 2024-10-18) Zou, Mengbang; Perrusquia, Adolfo; Guo, Weisi
    Explaining what does a data-driven control algorithm learns play a crucial role for safety critical control of autonomous platforms in transportation. This is more acute in reinforcement learning control algorithms, where the learned control policy depends on various factors that are hidden within the data. Explainable artificial intelligence methods have been used to explain the outcomes of machine learning methods by analysing input-output relations. However, data-driven control does not pose a simple input-output mapping and hence, the resulting explanations lack depth. To deal with this issue, this paper proposes a explainable data-driven control method that allows to understand what the data-driven method is learning from the data. The model is composed by a Q-learning algorithm enhanced by a dynamic mode decomposition with control (DMDc) algorithm for state-transition function estimation. Both the Q-learning and DMDc provides the elements that are learned from the data and allow the construction of counterfactual explanations. The proposed approach is robust and does not require hyperparameter tuning. Simulation experiments are conducted to observe the benefits and challenges of the method.
  • Thumbnail Image
    ItemOpen Access
    Infection and re-infection: stability of complex air transport network
    (IEEE, 2021-10-15) Zou, Mengbang; Guo, Weisi
    Global epidemic propagation rate and structure are strongly coupled with the international air transport network. Due to the network structure, countries are desynchronized in the way infections arise, transported, mitigated, and re-infect again. A global lockdown is detrimental to the international economy, and many argue that unless the whole world is evenly vaccinated, we cannot return to pre-COVID lives. The current challenge is that new waves of re-infection are spreading, and vaccination will take many months to materialise across the world. Here, we show how certain small airports (0.1% of global) contribute much more to the epidemic spread process, irrespective of the actual spreading mechanics. We use trophic coherence as a metric for network stability in directed graphs (e.g., recurring network source of reinfection). We find that the air transport network has a trophic coherence similar to a random expectation (99% of airports form a tightly looped network) and practically all networks need to be equally protected to ensure global security.
  • Thumbnail Image
    ItemOpen Access
    Local assortativity affects the synchronizability of scale-free network
    (IEEE, 2022-11-14) Zou, Mengbang; Guo, Weisi
    Synchronization is critical for system-level behavior in physical, chemical, biological, and social systems. Empirical evidence has shown that the network topology strongly impacts the synchronizability of the system, and the analysis of their relationship remains an open challenge. We know that the eigenvalue distribution determines a network's synchronizability, but analytical expressions that connect network topology and all relevant eigenvalues (e.g., the extreme values) remain elusive. Here, we accurately determine its synchronizability by proposing an analytical method to estimate the extreme eigenvalues using perturbation theory. Our analytical method exposes the role that global and local topology combine to influence synchronizability. We show that the smallest nonzero eigenvalue λ(2) , which determines synchronizability, is estimated by the smallest degree augmented by the inverse degree difference in the least connected nodes. From this, we can conclude that there exists a clear negative relationship between λ(2) and the local assortativity of nodes with the smallest degree value. We validate the accuracy of our framework within the setting of a scale-free network and can be driven by commonly used ordinary differential equations (e.g., 3-D Rosler dynamics or Hindmarsh–Rose neuronal circuit). From the results, we demonstrate that the synchronizability of the network can be tuned by rewiring the connections of these particular nodes while maintaining the general degree profile of the network.
  • Thumbnail Image
    ItemOpen Access
    Rewiring complex networks to achieve cluster synchronization using graph convolution networks with reinforcement learning
    (IEEE, 2024-06-10) Zou, Mengbang; Guo, Weisi; Chu, Kai-Fung
    Synchronization on complex networks depends on a myriad of factors such as embedded dynamics, initial conditions, network structure, etc. Current literature simplifies analysis of cluster synchronization leveraging conditions on network topology such as input-equivalence, network symmetries, etc., of which external equitable partition (EEP) is one of the most relaxed conditions. One practical problem is that for a dynamic system, how to alter a network to reach arbitrary achievable cluster synchronization and remaining faithful to the original structure. To solve this problem, we represent graph dynamics in Graph Convolution Network (GCN) modules that sit within an Actor-Critic Reinforcement Learning (AC-RL) framework under the condition of EEP. This allows the framework to select a good policy to sequentially rewire the network, where the sequence of moves matters. We test our method on two types of high-dimensional networked systems, Rossler dynamic networks and Hindmarsh-Rose neuronal circuits, with different network sizes. Our research opens up a way for the discovery of achievable cluster synchronization configurations by altering the network structure in any given networked dynamics.
  • Thumbnail Image
    ItemOpen Access
    Robust time synchronisation for industrial internet of things by H∞ output feedback control
    (IEEE, 2022-01-20) Zong, Yan; Dai, Xuewu; Wei, Zhuangkun; Zou, Mengbang; Guo, Weisi; Gao, Zhiwei
    Precise timing over timestamped packet exchange communication is an enabling technology in the mission-critical industrial Internet of Things, particularly when satellite-based timing is unavailable. The main challenge is to ensure timing accuracy when the clock synchronisation system is subject to disturbances caused by the drifting frequency, time-varying delay, jitter, and timestamping uncertainty. In this work, a Robust Packet-Coupled Oscillators (R-PkCOs) protocol is proposed to reduce the effects of perturbations manifested in the drifting clock, timestamping uncertainty and delays. First, in the spanning tree clock topology, time synchronisation between an arbitrary pair of clocks is modelled as a state-space model, where clock states are coupled with each other by one-way timestamped packet exchange (referred to as packet coupling), and the impacts of both drifting frequency and delays are modelled as disturbances. A static output controller is adopted to adjust the drifting clock. The H∞ robust control design solution is proposed to guarantee that the ratio between the modulus of synchronisation precision and the magnitude of the disturbances is always less than a given value. Therefore, the proposed time synchronisation protocol is robust against the disturbances, which means that the impacts of drifting frequency and delays on the synchronisation accuracy are limited. The one-hour experimental results demonstrate that the proposed R-PkCOs protocol can realise time synchronisation with the precision of six microseconds in a 21-node IEEE 802.15.4 network. This work has widespread impacts in the process automation of automotive, mining, oil and gas industries.
  • Thumbnail Image
    ItemOpen Access
    Uncertainty of resilience in complex networks with nonlinear dynamics
    (IEEE, 2020-11-24) Moutsinas, Giannis; Zou, Mengbang; Guo, Weisi
    Resilience is a system’s ability to maintain its function when perturbations and errors occur. Whilst we understand low-dimensional networked systems’s behavior well, our understanding of systems consisting of a large number of components is limited. Recent research in predicting the network level resilience pattern has advanced our understanding of the coupling relationship between global network topology and local nonlinear component dynamics. However, when there is uncertainty in the model parameters, our understanding of how this translates to uncertainty in resilience is unclear for a large-scale networked system. Here we develop a polynomial chaos expansion method to estimate the resilience for a wide range of uncertainty distributions. By applying this method to case studies, we not only reveal the general resilience distribution with respect to the topology and dynamics submodels but also identify critical aspects to inform better monitoring to reduce uncertainty.
  • Thumbnail Image
    ItemOpen Access
    Uncertainty quantification of multi-scale resilience in networked systems with nonlinear dynamics using arbitrary polynomial chaos
    (Nature Research, 2023-01-10) Zou, Mengbang; Zanotti Fragonara, Luca; Qiu, Song; Guo, Weisi
    Complex systems derive sophisticated behavioral dynamics by connecting individual component dynamics via a complex network. The resilience of complex systems is a critical ability to regain desirable behavior after perturbations. In the past years, our understanding of large-scale networked resilience is largely confined to proprietary agent-based simulations or topological analysis of graphs. However, we know the dynamics and topology both matter and the impact of model uncertainty of the system remains unsolved, especially on individual nodes. In order to quantify the effect of uncertainty on resilience across the network resolutions (from macro-scale network statistics to individual node dynamics), we employ an arbitrary polynomial chaos (aPC) expansion method to identify the probability of a node in losing its resilience and how the different model parameters contribute to this risk on a single node. We test this using both a generic networked bi-stable system and also established ecological and work force commuter network dynamics to demonstrate applicability. This framework will aid practitioners to both understand macro-scale behavior and make micro-scale interventions.