performance, Multi-objective optimization.

Stochastic nonlinear and Multi-Agent Systems (MASs) are increasingly prevalent in modern engineering applications, from autonomous vehicles to smart grids, but their inherent uncertainties and complex interactions pose significant control challenges. Fuzzy control has emerged as an essential solution, offering robust handling of nonlinearities, approximation of unknown dynamics, and adaptive decision-making under uncertainty.

Recent advances have focused on four key themes: consensus control for coordinated MAS behavior, event-triggered control to reduce communication overhead, prescribed-performance and finite/fixed-time tracking to ensure guaranteed convergence, and optimization-oriented approaches that integrate fuzzy logic with adaptive and multi-objective algorithms.

This review aims to synthesize these developments, provide a critical comparison of existing methodologies, identify research gaps, and suggest future directions. The paper is organized as follows: Section 2 discusses fuzzy consensus strategies, Section 3 covers event-triggered and switching control methods, Section 4 addresses prescribed-performance and finite-time tracking, Section 5 reviews optimization-based fuzzy approaches, Section 6 highlights emerging trends and challenges, and Section 7 concludes with insights and future research directions

The study of fuzzy control for stochastic nonlinear and MASs is rooted in the fusion of fuzzy logic theory, nonlinear control, and probabilistic modeling. Traditional control methods often rely on precise mathematical models; however, real-world systems such as autonomous vehicles, sensor networks, and robotic swarms exhibit uncertainties, nonlinearities, and random disturbances that make exact modeling infeasible. Fuzzy logic, introduced by Zadeh [1], provides a framework for handling linguistic uncertainty and imprecision, enabling controllers to approximate human reasoning and adapt to uncertain environments.

In stochastic nonlinear systems, uncertainties arise both from system dynamics and random noise. Fuzzy adaptive control integrates learning mechanisms that adjust control rules in real time to maintain stability and performance despite such disturbances. When extended to MASs, fuzzy control supports distributed decision-making, allowing agents to achieve consensus or cooperation under communication constraints and environmental uncertainties.

Key theoretical tools include Lyapunov stability theory, Itô stochastic calculus, and fuzzy approximation lemmas, which guarantee convergence and robustness. The development of finite-time, fixed-time, and prescribed-performance control frameworks further enhances system responsiveness and reliability. Recent extensions incorporate event-triggered control to reduce communication burden and fuzzy optimization techniques such as Adaptive Dynamic Programming (ADP) and chance-constrained programming to achieve energy-efficient and resilient coordination.

Overall, this theoretical foundation establishes fuzzy control as a vital approach for managing the nonlinearity, randomness, and decentralization inherent in complex stochastic MASs, paving the way for advanced applications in intelligent robotics, networked systems, and cyber-physical environments.

Over the past decade, the fusion of fuzzy logic, stochastic systems theory, and distributed optimization has produced a rich body of methods for controlling nonlinear and networked systems under uncertainty. Foundational advances in Takagi–Sugeno fuzzy modeling and stochastic Lyapunov analysis enabled fuzzy adaptive controllers to approximate unknown nonlinear dynamics while guaranteeing mean-square or probabilistic stability. Building on those foundations, recent research (2020–2025) has concentrated on four convergent themes: consensus and cooperative control, event-triggered communications, prescribed-performance and finite/fixed-time stability, and fuzzy-guided optimization. The following review synthesizes these strands and highlights representative contributions and open issues.

Consensus and cooperative control for stochastic MASs have evolved from model-based distributed controllers to adaptive fuzzy frameworks that tolerate uncertainties, topology switching, and partial faults. Chen et al. [2] propose a robust fuzzy cooperative strategy to achieve global consensus in stochastic MASs and demonstrate resilience to stochastic perturbations. Li & Li [3] extend fuzzy consensus to include fault tolerance and optimality considerations, while Niu et al. [4] and Yi et al. [5] present fixed- and prescribed-time consensus techniques that preserve connectivity under switching dynamics and constrained interactions. Collectively, these works show that fuzzy approximators can compensate for nonlinearities across agents and that Laplacian-based consensus formulations remain effective when coupled with adaptive fuzzy laws. However, scalability to large heterogeneous networks and energy-aware communication scheduling remain underexplored in the literature [2–5].

Event-triggered control has been widely adopted to mitigate communication overhead in networked MASs. Zhang et al. [6] and Wu et al. [7] develop adaptive and state-dependent triggering rules that guarantee finite-time convergence for nonlinear stochastic systems; Xu et al. [8] further propose dynamic event-triggered mechanisms tailored to switching topologies and optimization objectives. Notably, Zhang et al. [6] incorporate measurement sensitivity and deception-attack models, marking a shift toward cyber-resilient event-triggered designs. Despite these advances, most triggering schemes require conservative bounds or prior knowledge of system norms, and learning-based adaptive triggering (e.g., Reinforcement Learning (RL)- driven thresholds) and rigorous analysis of the triggering–stability trade-off under rapidly varying noise remain open problems [6–9].

Prescribed-performance control and finite/fixed-time stabilization provide explicit transient and steady-state guarantees critical for safety-critical applications. Sun et al. [10], Fu et al. [11], and Xu et al. [12] introduce fuzzy adaptive switching or fixed-time controllers for stochastic high-order and non-triangular systems, employing error transformation techniques to enforce performance funnels. These approaches achieve bounded transient behavior and rapid convergence, but frequently assume ideal actuators or neglect actuator saturation and communication delays; extending prescribed performance to chance-constrained stochastic settings would better capture probabilistic risk measures in real environments [5], [10], [11], [12].

Optimization-driven fuzzy control has attracted increasing attention. Zhang et al. [9] and Zheng et al. [13] integrate ADP and distributed adaptive optimization with fuzzy approximators to obtain near-optimal consensus policies under uncertainty and delays. Wang et al. [14] and Kumar & Mishra [15] explore fuzzy-guided multi-objective and chance-constrained optimization methods, indicating potential for balancing competing objectives (robustness, energy, convergence speed). The main challenge here is computational tractability: ADP and stochastic chance-constrained formulations often impose heavy online computation, motivating the development of distributed, lightweight solvers or surrogate models that maintain theoretical guarantees [9], [13–15].

Several cross-cutting themes deserve emphasis. First, a trend toward hybrid solutions, combining fuzzy approximators with ADP, RL, or metaheuristics, is apparent, but few works provide rigorous stochastic stability proofs for these hybrids. Second, security-aware control (e.g., deception attacks) is nascent but crucial given practical cyber-physical threats [6]. Third, most contributions validate methods via simulations; hardware-in-the-loop and real-world experiments (robotic swarms, microgrids, sensor networks) are sporadic and are a necessary next step for demonstrating robustness and real-time feasibility.

In summary, the recent literature (2020–2025) demonstrates that fuzzy control techniques can achieve robust consensus, efficient event-triggering, prescribed performance, and optimization in stochastic nonlinear MASs. The primary future directions are the development of scalable distributed algorithms, learning-augmented event-triggering and ADP, probabilistic/prescribed-performance fusion, and experimental validation across application domains such as robotics, energy systems, and process control.

Zhang et al. Event-triggered adaptive fuzzy attack models

Wu et al. [7] Finite-time adaptive fuzzy switching (event-triggered) Xu et al. [8] Dynamic event-triggered fuzzy optimal consensus Zhang et al. ADP + adaptive fuzzy; event-triggered

Wang et al. Hierarchical fuzzy-guided multi-objective [14] algorithm Kumar and Nonlinear chance-constrained fuzzy Mishra [15] optimization Li and Li [3] Fault-tolerant fuzzy optimal consensus

Yang et al. H∞ fuzzy switching adaptive control [16] Yi et al. [5] Prescribed-time connectivity-preserving consensus Zheng et al. Distributed adaptive fuzzy optimization [13] Niu et al. [4] Nonsingular fixed-time bipartite consensus

Xu et al. [12] Adaptive fuzzy fixed-time control

Fu et al. [11] Finite-time fuzzy tracking with PPC

optimization

adaptation

from data

Main Modeling/Assumption(S) Representative Application(S) Stochastic MAS; known/partially Multi-robot consensus, formation known graph control Itô SDEs; finite-time bounds fixed Process control, high-order systems offline Measurement attacks; stochastic noise Secure sensor networks, CPS

Non-affine dynamics; bounded noise Nonlinear control, autonomous agents

Switching topology; dynamic Networked control, distributed thresholds estimation Delays; unknown dynamics; ADP Optimized MAS coordination, UAV approximations swarms Dynamic multi-objective formulation Energy management, scheduling

Probabilistic constraints; mixed Power systems, resource allocation fuzzy-stochastic Fault models; stochastic disturbances Fault-resilient MASs

Markovian hybrid switching; H∞ Hybrid process systems metric Switched MASs; connectivity Robotic swarms, formation constraints maintenance Partial observability; distributed info Distributed control, microgrids

Constrained MAS; power integration Constrained networks, constrained technique robotics Non-triangular stochastic switching Robust nonlinear control systems High-order stochastic systems Precision tracking, industrial control

efficiently

online use

across hierarchy

verification under switching

guarantees Fuzzy + Neural nets (neuro-fuzzy) Flexible approximation, ability to learn Interpretability loss; need for provable

Recent advances in adaptive fuzzy control for stochastic nonlinear and MASs have focused on improving consensus, stability, efficiency, and resilience under uncertainty. Studies such as Chen et al. [2] and Li & Li [3] developed robust and fault-tolerant fuzzy consensus strategies, while Xu et al. [8] and Zhang et al. [9] extended these methods to switching topologies and delayed networks using event-triggered mechanisms to reduce communication overhead. Event-triggered and finite-/fixed-time control approaches, as seen in Zhang et al. [6], Wu et al. [7], and Niu et al. [4], guarantee rapid convergence and prescribed performance, ensuring system states meet predefined transient and steady-state requirements. In parallel, fuzzy-guided optimization techniques, highlighted by Wang et al. [14], Kumar & Mishra [15], and Zheng et al. [13], have addressed multi-objective stochastic optimization problems, providing systematic uncertainty handling and improved control efficiency. A few studies, including Zhang et al. [6] and Yang et al. [16], also considered security and resilience against measurement attacks. Despite these innovations, current approaches face limitations such as high computational complexity, limited scalability to large-scale networks, insufficient integration of consensus, event-triggering, and optimization, and inadequate security-aware designs, with most validation performed in simulation rather than practical deployments. These gaps highlight opportunities for developing efficient, unified, and resilient fuzzy control frameworks for complex stochastic systems.

A comprehensive comparative evaluation of fuzzy-adaptive and hybrid control methods reveals varying strengths across key performance metrics. In terms of convergence, finite-time and fixed-time control designs ([4], [6], [7], [10–12]) demonstrate the fastest error decay, making them ideal for safety-critical applications. However, these approaches often demand conservative gains or nonlinear structures that increase tuning complexity and energy cost. Optimization-driven techniques such as ADP [9], [13] offer near-optimal transient performance but lack strict finite-time guarantees without additional design refinements. Regarding computational cost, classical adaptive fuzzy and event-triggered controllers [2], [6], [7] remain lightweight and practical, while ADP-based, chance-constrained, and hierarchical fuzzy-optimization frameworks [9], [14], [15] achieve superior performance at the expense of higher computational and memory burdens, which limit their real-time applicability. In robustness, methods incorporating fault models or H∞ metrics [2–4], [6], [16] exhibit stronger disturbance rejection, whereas ADP-based controllers may become sensitive to modeling errors and time delays. Scalability remains a challenge, as most designs are validated only on small-scale systems; although distributed and hierarchical fuzzy structures [13], [14] show promise, communication and consensus costs must be effectively managed. Adaptability is a clear advantage of fuzzy-adaptive designs [2– 4], [7], [13], which excel in real-time parameter adjustment, while learning-based hybrids enhance flexibility in dynamic environments but at notable computational cost.

From an individual method perspective, pure fuzzy-adaptive controllers provide interpretability, strong nonlinear approximation, and provable stability, but face scalability and rule-based dependency issues. Event-triggered fuzzy control offers communication efficiency and finite-time convergence yet requires careful threshold design to avoid Zeno behavior. Prescribed-performance and finite/fixed-time fuzzy controllers ensure rapid convergence but may impose high control effort and rely on idealized assumptions. Optimization-based fuzzy controllers, including ADP and multi-objective designs, balance competing objectives effectively but suffer from heavy computation and potential sensitivity to initialization. Fault-tolerant and H∞ fuzzy switching controllers deliver excellent disturbance rejection, though often at the cost of conservatism.

In hybrid frameworks, Fuzzy–ADP/RL integration merges interpretability with data-driven optimality, enhancing adaptability in uncertain systems but struggling with computational intensity and stochastic stability guarantees. Fuzzy–Chance-Constrained optimization effectively handles probabilistic risk in stochastic environments but faces nonconvex constraint challenges and reliance on sampling approximations. Fuzzy– Metaheuristic combinations (GA/PSO) enable global search and multi-objective tuning but remain largely offline, limiting real-time responsiveness. Lastly, Fuzzy–hierarchical or distributed control structures enhance modularity and large-scale coordination, though inter-layer stability analysis and communication overhead continue to pose significant challenges.

Together, these findings underscore the trade-offs between performance, complexity, and scalability, highlighting the need for future designs that blend fuzzy interpretability, learning-based adaptability, and computational efficiency within distributed, real-time architectures.

Despite significant progress in fuzzy-adaptive control for MASs, several research gaps persist across key thematic areas. In consensus and cooperative control [2–5], [8], [9], robust fuzzy cooperative strategies have successfully addressed stochastic uncertainties, switching topologies, fault tolerance, and finite/fixed-time connectivity-preserving consensus. However, most studies assume ideal or partially known communication graphs, leaving dynamic topology uncertainties and random communication failures largely unaddressed. Scalability to large-scale heterogeneous MASs and optimization of energy and communication resources remain insufficiently explored, and real-time adaptability under rapidly varying environmental noise is still limited.

Event-triggered and switching control [6–9] has reduced communication burdens and enabled adaptive triggering laws ensuring finite/fixed-time convergence, with some efforts incorporating security through deception attack models. Nevertheless, trade-offs between triggering frequency and stability margins under stochastic disturbances are rarely analyzed. Existing triggering conditions rely on known system bounds, limiting applicability to unknown or time-varying nonlinear systems, and data-driven or learning-based adaptive triggering remains underdeveloped. Cybersecurity-aware event-triggering schemes resilient to coordinated attacks are also largely absent.

Prescribed-performance and finite/fixed-time control [4], [5], [10–12], [16] have achieved bounded-error guarantees, rapid stabilization using Adding Power Integration (API), and adaptive law tuning for high-order systems. Yet, unified frameworks linking prescribed performance with probabilistic reliability or chance constraints are lacking. Many finite/fixed-time controllers assume ideal actuator dynamics, ignoring saturation and time delays, while the quantitative trade-off between convergence speed and control effort is underexplored. Performance guarantees under hybrid switching or Markovian uncertainties are also limited. Fuzzy-guided optimization and learning approaches [9], [13–15] have applied fuzzy-ADP for policy optimization under stochastic MASs, integrated multi-objective and chance-constrained formulations, and explored hierarchical fuzzy systems for dynamic optimization. However, most approaches remain centralized, with distributed optimization under stochastic settings still in early development. Computational scalability and real-time implementability of fuzzy-ADP or multi-objective fuzzy algorithms are not fully demonstrated, benchmark datasets and standardized performance metrics are scarce, and integration with RL or metaheuristics for online adaptation is limited.

Across these domains, cross-thematic integration gaps persist. Event-triggered control has rarely been combined with fuzzy-guided optimization under prescribed-performance frameworks. Security and resilience against cyber-physical threats are seldom integrated with finite-time stability guarantees. Standardized frameworks or simulation benchmarks for comparing different fuzzy-control paradigms in stochastic nonlinear systems are largely absent, and most studies remain theoretical, with minimal hardware-in-loop or experimental validation. These gaps highlight opportunities for future research to develop scalable, resilient, and experimentally validated fuzzy-adaptive control architectures that unify performance, robustness, and real-time adaptability.

The reviewed studies from 2020–2025 reveal significant progress in adaptive fuzzy control for stochastic nonlinear and MASs, yet several open directions remain for future exploration. First, while recent advances [2], [8] successfully integrate event-triggered and fuzzy-guided optimization, there is still a need to develop scalable architectures that can efficiently handle large-scale heterogeneous MASs with limited communication and computation resources. Future research should also focus on hybrid learning-based fuzzy frameworks, combining RL, neural adaptive approximators, and model-free strategies to enhance generalization under high-dimensional uncertainties. Moreover, despite notable progress in finite-time and fixed-time convergence [4], [12], the trade-off between convergence speed, robustness, and energy efficiency requires systematic investigation, especially under switching or uncertain topologies.

Another promising direction lies in secure and resilient cooperative control, where the impacts of cyber-attacks, deception, and communication delays [6], [7] need to be counteracted using distributed event-triggered fault-tolerant schemes. Similarly, chance-constrained and fuzzy-stochastic optimization [15] offer fertile ground for multi-objective control under probabilistic risk measures, but require improved computational efficiency through decentralized and data-driven algorithms. In addition, prescribed-performance control [5], [11] should be extended to non-strict-feedback and non-triangular system structures, addressing real-world dynamics beyond idealized models. Finally, multi-layer fuzzy cooperative learning, incorporating hierarchical decision-making and adaptive distributed optimization [13], [14], can pave the way toward autonomous, self-evolving intelligent MASs that achieve robustness, adaptability, and sustainability in uncertain stochastic environments.

Emerging directions in fuzzy-adaptive control: the next generation of fuzzy-adaptive control for MASs is driven by the need for scalability, resilience, and real-time adaptability across complex networked environments. In consensus and cooperative fuzzy control ([1], [5], [6], [9], [11], [13]), research should move toward scalable, distributed architectures capable of handling large, heterogeneous agent networks with varying dynamics, communication delays, and learning capabilities. Resilient consensus frameworks must integrate Markovian jump processes, time delays, and packet losses for realistic network modeling, while energy-efficient communication scheduling can balance accuracy with resource consumption. Moreover, human-in-the-loop cooperative designs are essential for cyber–physical–human systems where human feedback influences collective decision-making.

Event-triggered and switching mechanisms require data-driven and learning-based triggering laws, leveraging RL or neural-fuzzy systems to adapt event thresholds in real time [6–9]. Cyber security considerations, including intrusion detection and deception-resilient triggering, must be incorporated, alongside hierarchical and decentralized multi-layer designs that reduce computation and prevent Zeno behavior. Hardware-implementable solutions are needed to validate real-time feasibility under sampling noise and embedded constraints.

Prescribed-performance and finite/fixed-time control [4], [5], [10–12], [16] should evolve toward unified stochastic frameworks, integrating probabilistic and chance-constrained formulations. Actuator-aware finite-time controllers must respect saturation and safety constraints, while explicit trade-offs between energy consumption, convergence speed, and noise resilience should be quantified. Adaptive neural–fuzzy hybrid designs can enable real-time parameter tuning, and multi-objective frameworks should simultaneously ensure stability, accuracy, energy efficiency, and fault tolerance.

Fuzzy-guided and optimization-based control [9], [13–15] is poised to benefit from distributed architectures that scale across partially observable networks. Hybridization with evolutionary and swarm algorithms (GA, PSO, differential evolution) can enhance multi-objective optimization, while improvements in computational efficiency and convergence guarantees are necessary for real-world deployment. Standardized benchmarks are critical for fair algorithmic comparison, and interpretable fuzzy-ADP designs will maintain explainability alongside performance.

Cross-thematic integration represents the emerging frontier. Future research should develop unified frameworks combining consensus, event-triggered control, and prescribed performance under fuzzy optimization, incorporating uncertainty quantification and Bayesian fuzzy modeling for robust decision-making. Digital twin–based MAS control, coupled with AI-driven perception and self-learning fuzzy rules, can enable adaptive prediction and autonomous decision-making. Finally, experimental validation in robotic swarms, autonomous vehicles, and sensor networks will be essential to bridge the gap between theoretical developments and practical deployment.

protocols, and resilient communication in uncertain large-scale networks Event-triggered control RL-driven triggering, cyber-resilient designs, multi- Efficient communication and layer coordination enhanced security Prescribed Performance Chance-constrained bounds, constraint-aware Guaranteed performance under design, adaptive hybrid fuzzy-RL uncertainty Optimization Distributed fuzzy-ADP, explainable RL, Real-time, interpretable, scalable benchmark frameworks optimization Cross-integration Unified hybrid frameworks, digital twins, Practical, AI-integrated fuzzy experimental deployment control systems adaptation, scalability, and security.

Key Limitations Identified Future Research Opportunities Limited scalability and computational Develop distributed, scalable fuzzy complexity under large agent networks consensus strategies with adaptive communication and energy-efficient protocols Performance bounds fixed offline; Design online self-tuning lacks adaptability to changing performance bounds using learning-environments based fuzzy inference Vulnerable to complex cyber threats; Develop secure, intrusion-resilient limited experimental validation event-triggered fuzzy mechanisms with anomaly detection Focuses on single-agent systems; Extend to distributed multi-agent limited generalization to MASs architectures and hybrid switching topologies Switching logic increases instability Optimize triggering thresholds and control cost using RL-based fuzzy adaptation for stability–efficiency balance High computational load and Develop simplified or distributed sensitivity to delay ADP frameworks using parallel fuzzy learning.

Limited interpretability and real-time Create explainable fuzzy–neural feasibility optimization methods for dynamic systems Complex probabilistic modeling; Integrate lightweight surrogate computationally expensive models and online adaptive uncertainty quantification Assumes ideal communication; limited Embed predictive fault detection fault diagnosis capability and recovery in distributed fuzzy control loops Conservative stability conditions; lacks Extend to stochastic hybrid systems hybrid uncertainty handling with variable mode-dependent uncertainties Restricted to switched topologies with Address dynamic and time-varying fixed connectivity network reconfiguration in real time Scalability limited; convergence rate Develop hierarchical distributed not guaranteed fuzzy optimization with convergence–accuracy trade-offs High control energy and complex Design energy-efficient, adaptive tuning fixed-time control with adjustable exponents.

Focused on specific non-triangular Generalize to arbitrary nonlinear structures dynamics and uncertain switching networks.

Limited robustness to unknown Combine fuzzy–neural estimation disturbances with robust optimization for disturbance rejection.

Across all references, there is a clear transition from model-based fuzzy control to data-driven, distributed, and intelligent fuzzy systems. The next generation of research should unify consensus, event-triggering, prescribed performance, and optimization into self-learning fuzzy frameworks capable of real-time

The reviewed studies collectively demonstrate that fuzzy control has evolved into a versatile framework for addressing the inherent nonlinearity, stochasticity, and inter-agent coupling in modern MASs. Through diverse design paradigms, ranging from finite- and fixed-time consensus to event-triggered communication, prescribed-performance regulation, and fuzzy-guided optimization, researchers have substantially enhanced convergence rates, disturbance rejection, and adaptability under uncertain and dynamic environments. In particular, finite-time and fixed-time controllers exhibit superior transient behavior, ensuring rapid coordination, while event-triggered mechanisms effectively balance control performance with communication efficiency. Similarly, optimization-driven fuzzy schemes based on ADP or chance-constrained formulations provide near-optimal coordination and robust decision-making under probabilistic uncertainty. However, the computational complexity, scalability limitations, and incomplete stochastic stability proofs still hinder their large-scale deployment. Most current works focus on moderate-sized MASs and deterministic conditions, leaving open challenges in scalability, cybersecurity resilience, delay compensation, and distributed learning-based adaptation.

Looking forward, future research should aim at integrating fuzzy intelligence with data-driven RL and distributed optimization to enable real-time adaptive control with provable stability in high-dimensional stochastic networks. Establishing unified evaluation metrics, including convergence speed, communication load, and robustness under uncertainty, will also be critical for objective benchmarking. Moreover, the development of energy-efficient and secure event-triggered strategies that remain effective under network attacks or communication faults represents a vital direction for next-generation intelligent MAS coordination. Overall, the synthesis of fuzzy logic, optimization theory, and distributed learning is expected to bridge the gap between theoretical design and practical deployment, paving the way toward scalable, resilient, and self-evolving control architectures for complex stochastic MASs.

The author wishes to acknowledge the extensive body of literature and scholarly work that served as the basis for this review article. Sincere thanks are also extended to colleagues, mentors, and peers for their insightful suggestions, critical feedback, and continuous encouragement during the conceptualization and writing of this manuscript. The author alone is responsible for the interpretations and any errors that may appear in the paper.

This review article was conceptualized, written, and finalized solely by the author.

The author did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors for this review.

This review article is based solely on previously published literature. No original datasets were generated or used, and therefore no data are available for sharing.

The author declares that there is no conflict of interest regarding the publication of this review article.

This article does not involve any studies with human participants or animals conducted by the author. Therefore, ethical approval was not required.