Human-on-the-Loop Distributed Control of Grid-Interactive Buildings for Demand Response Participation
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.
Rosa: We're looking at the paper: "Human-on-the-Loop Distributed Control of Grid-Interactive Buildings for Demand Response Participation".
Dev: Here's the gist of the paper. The paper proposes a human-in-the-loop distributed consensus control approach for demand-side management across multiple buildings specifically introducing a novel framework where a human acts as the non-autonomous leader in consensus control of cooperative buildings participating in demand response programs and this system enables flexible management of energy use and thermal comfort during demand response events.
Rosa: System Modeling and Problem Statement The system is modeled as a leader-follower MAS where cooperative buildings aim to track the state of the facility building, which receives DR signals from the utility and the facility manager determines participation based on operating conditions and occupant comfort.
Dev: The core challenge involves incorporating a human as a leader, which introduces complexities due to variability and nonlinear responses, posing difficulties in achieving reliable control.
Rosa: Specifically, challenges include unknown exogenous inputs from the facility manager's decisions that reflect trade-offs among thermal comfort and grid service provision, heterogeneous thermal dynamics requiring strategies respecting diverse physical limits, communication constraints necessitating distributed architectures, and thermal models violating assumptions for standard unknown input observers.
Title and authors: Rosa: So let’s talk about who wrote this, and what this paper, "Human-on-the-Loop Distributed Control of Grid-Interactive Buildings for Demand Response Participation," actually proposes.
Dev: The authors are Kasra Mazarei Saadabadi, Dongming Wang, Wei Ren, Alfredo Martinez Morales, and Hamidreza Nazaripouya. They’re all senior members in IEEE and they’ve put together this control approach.
Rosa: What I find interesting is that the paper introduces a specific structure where the facility manager in the main building serves as the leader to guide the others.
Dev: It sets up a leader-follower multi-agent system, which is standard for coordinated behavior, but they’re adding that human element as a nonautonomous input.
Taro: From my side, I see this immediately as a way to introduce real uncertainty into the control loop; the human decisions are unpredictable and really reflect trade-offs between comfort and grid service.
Rosa: That’s right, and their system is modeled in such a way that the cooperative buildings try to track the state of that facility building, which is getting signals from the utility.
Dev: They have a specific problem with how they estimate those states because some building states can't be measured directly in practice, so they need some estimation methods.
Taro: That’s where their proposed method gets interesting; they suggest using a nonlinear unknown input sliding mode observer, specifically tailored for these leader-follower multiagent systems.
The paper's summary: Rosa: So what the paper summarizes is that to handle those unmeasurable states, they propose a sliding mode observer because the standard observer matching condition isn't satisfied in this context.
Dev: That’s because the unknown input—which comes from the human leader’s decisions—isn't directly reflected in what you measure, so they use a sliding mode to enforce a sliding manifold on the estimation error dynamics.
Taro: It sounds robust; it means that even when those human decisions are erratic or unpredictable, the observer can still keep up with estimating what’s going on.
Rosa: And they focus this distributed observer design specifically on estimating the consensus error using local measurements and information from neighboring buildings to estimate that error.
Dev: They’re trying to reduce communication overhead too, so each building only shares information with its immediate neighbors instead of everyone in the network.
Taro: It’s smart to limit communication, but it makes estimation harder when you don't have all the data points available across the whole system.
The paper's improvements: Rosa: Now let’s look at what they suggest as improvements to this control approach. The paper mentions using a leader-follower consensus protocol where the nominal control input uses a saturation function to respect physical actuator limits.
Dev: They define that pre-saturation control signal as v sub i equals negative K i delta i minus rho i sgn(K i delta i), and the consensus error delta sub i is defined by the interaction between neighbor states and the leader state.
Taro: That saturation function part is important because it’s how they make sure the buildings actually stay within their physical limits when they are trying to follow that leader’s plan.
Rosa: They argue this protocol ensures that each building adjusts its state appropriately, aligning the entire network with the facility manager’s signal as the leader.
Dev: So, it's not just about following a command; it's about coordinating participation in demand response events while simultaneously maintaining indoor temperatures within an acceptable range and ensuring consistent comfort levels across all participating buildings.
Conclusion: Rosa: So to wrap up this study on "Human-on-the-Loop Distributed Control of Grid-Interactive Buildings for Demand Response Participation," they show that stability is achieved under Assumption five which involves structured stabilizing feedback and self consistent design tuples.
Dev: The analysis shows that if assumptions one through five hold and the decay margin condition is satisfied, then the Lyapunov function is forward invariant and satisfies an ultimate bound as time goes to infinity.
Taro: It means that even with all those uncertainties from the human leader, there’s a guaranteed way for the system to settle down into a stable residual set over time.
Rosa: In simulation validation, they tested this approach across three scenarios: baseline demand response, mid-course revision of the event, and delayed commitment.
Dev: Scenario A showed that they successfully reduced demand by three hundred kilowatt while maintaining a new comfort state setpoint.
Taro: Scenario B was particularly interesting because it demonstrated robust consensus tracking control under abrupt changes in human control input by showing buildings quickly realign with the leader’s updated state after a revision.
Rosa: And Scenario C illustrated that there’s a trade-off between grid support and comfort, showing that indoor temperatures can exceed the nominal comfort range near the end of an aggressive curtailment event.
Dev: The simulation results confirm that all those transients remain bounded and decay rapidly, which validates the observer performance under sudden operating changes.
Taro: So what this means for us is that integrating human decision-making into advanced control systems can really enhance flexibility and adaptability in demand response by seamlessly incorporating human judgment into the control architecture.
Rosa: It definitely paves the way for further exploration of human-in-the-loop control strategies in other complex, dynamic power system applications.
Dev: The methodology successfully achieves consensus and maintains stability across various demand response scenarios, which is a solid result given the inherent complexity of having a human leader.
University of California Riverside · Oklahoma State University
eess.SY, cs.SY
Submitted: 2026-07-21
Updated: 2026-11-10
Comments: 10 pages, 9 figures
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 80/100
The gist: The paper proposes a human-in-the-loop distributed consensus control approach for demand-side management across multiple buildings specifically introducing a novel framework where a human acts as the…
Key concepts
- Human-in-the-Loop Leader
- A human facility manager acts as the non-autonomous leader in the system. This person decides how much energy demand each building should reduce based on trade-offs between keeping occupants comfortable and providing necessary grid services.
- Leader-Follower Consensus Control
- This is a control strategy where one building (the leader) dictates the desired state, and all other buildings (followers) adjust their operation to match it. This ensures all cooperative buildings work together toward a unified goal set by the human leader.
- Sliding-Mode Observer
- A specific estimation technique used to estimate internal system states when standard methods fail. It forces the estimation error dynamics onto a predefined 'sliding manifold,' allowing the system to accurately track states even when inputs are unknown or complex, like those from a human leader.
Terminology
Summary
The paper proposes a human-in-the-loop distributed consensus control approach for demand-side management across multiple buildings specifically introducing a novel framework where a human acts as the non-autonomous leader in consensus control of cooperative buildings participating in demand response programs and this system enables flexible management of energy use and thermal comfort during demand response events. p. 0
System Modeling and Problem Statement
The system is modeled as a leader-follower MAS where cooperative buildings aim to track the state of the facility building, which receives DR signals from the utility and the facility manager determines participation based on operating conditions and occupant comfort. The core challenge involves incorporating a human as a leader, which introduces complexities due to variability and nonlinear responses, posing difficulties in achieving reliable control. p. 0 Specifically, challenges include unknown exogenous inputs from the facility manager's decisions that reflect trade-offs among thermal comfort and grid service provision, heterogeneous thermal dynamics requiring strategies respecting diverse physical limits, communication constraints necessitating distributed architectures, and thermal models violating assumptions for standard unknown input observers.
Observer Design Methodology
To address the need for state estimation due to unmeasurable internal states like the derivative of the thermal comfort state, two primary observer approaches are considered: full-order distributed observers and interval observers. The paper notes that observer matching conditions, essential for standard linear unknown input observers, are not satisfied in this context because the unknown input is not directly reflected in the measured output. p. 2 Consequently, sliding-mode observers are motivated as an alternative because they enforce a sliding manifold on the estimation error dynamics, compensating for unknown inputs without requiring the matching condition. p. 4 The proposed observer design focuses on estimating consensus error using a distributed observer that processes local measurements and neighboring information to estimate this error. p. 3
Control Protocol Implementation
The control objective is to coordinate participation in DR events while maintaining indoor temperatures within an acceptable range and ensuring consistent comfort levels across all participating buildings. A leader-follower consensus protocol is employed where the nominal control input uses a saturation function to respect physical actuator limits. p. 3 The pre-saturation control signal is defined as vi = −Kiδi − ρi sgn(Kiδi), and the consensus error δi is defined by the interaction between neighbor states and the leader state. p. 3 The protocol ensures that each building adjusts its state appropriately, aligning the network with the facility manager’s signal as the leader. p. 3
Stability Analysis Results
The stability analysis employs a composite Lyapunov function combining per-agent and composite supertwisting Lyapunov functions to analyze the closed-loop dynamics. The analysis demonstrates that under Assumption 5, which involves structured stabilizing feedback and self-consistent design tuples, the system exhibits ultimate boundedness in the residual set. p. 5 Specifically, Theorem 3 shows that if Assumptions 1–5 hold and the decay-margin condition is satisfied, then V∗ is forward invariant, and the Lyapunov function satisfies an ultimate bound. p. 6 The final result indicates that the consensus error, observer error, and leader–follower tracking error satisfy specific bounds as time approaches infinity. p. 6
Simulation Validation
Simulation results validate the effectiveness of the proposed approach across three SCE-inspired DR scenarios: baseline DR, mid-course revision, and delayed commitment. Scenario A shows that the system successfully reduces demand by 300 kW while maintaining a new comfort state setpoint. p. 6 Scenario B demonstrates robust consensus tracking control under abrupt changes in human control input by showing buildings quickly realign with the leader’s updated state after a revision. p. 7 Scenario C illustrates the trade-off between grid support and comfort, showing that indoor temperatures can exceed the nominal comfort range near the end of an aggressive curtailment event. p. 8 The simulation confirms that all transients remain bounded and decay rapidly, confirming stable observer performance under sudden operating changes. p. 7
Conclusion
The proposed leader-follower consensus control ensures system stability while allowing the human leader to balance user thermal comfort and grid support. The results validate the effectiveness of integrating human decision-making into advanced control systems, enhancing flexibility and adaptability in DR by seamlessly incorporating human judgment into the control architecture. p. 9 This study paves the way for further exploration of human-in-the-loop control strategies in other complex, dynamic power system applications. p. 9 The proposed methodology successfully achieves consensus and maintains stability across various demand response scenarios. p. 9
ACKNOWLEDGMENT
This work is supported in part by CEC grant EPC-16-077.
REFERENCES
[1] A. M. Annaswamy and Y. p. 9 Yildiz, “Cyber-physical-human systems,” in Encyclopedia of systems and control pp 497–508, Springer, 2021; [3] Z. p. 9 Zhao, W. C. Lee, Y. Shin, and K. p. 9-B. Song, “An optimal power scheduling method for demand response in home energy management system,” IEEE Transactions on Smart Grid vol 4 no 3 pp 1391–1400, 2013; [8] P. p. 9 Palensky and D. p. 9 Dietrich, “Demand side management: Demand response, intelligent energy systems, and smart loads,” IEEE Transactions on Industrial Informatics vol 7 no 3 pp 381–388, 2011; [10] J. p. 9 Cai, J. E. Braun, D. Kim, and J. p. 9 Hu, “A multi-agent control based demand response strategy for multi-zone buildings,” in 2016 American Control Conference (ACC), 2016 pp 2365–2372; [18] P. p. 9 Wang, G. Wen, W. Yu, T. Huang, and X. p. 9 Yu, “An adaptive continuous approach to consensus tracking of nonlinear multiagent systems with a nonautonomous leader,” IEEE Transactions on Systems, Man, and Cybernetics: Systems vol 53 no 8 pp 4684–4695, 2023; [19] Y. p. 9 Hong, J. Hu, and L. Gao, “Tracking control for multiagent consensus with an active leader and variable topology,” Automatica vol 42 no 7 pp 1177–1182, 2006; [30] S. p. 9 Kundu, N. Sinitsyn, S. p. 9 Backhaus, and I. p. 9 Hiskens, “Modeling and control of thermostatically controlled loads,” 2011; [35] A. p. 9 Levant, “Robust exact differentiation via sliding mode technique,” Automatica vol 34 no 3 pp 379–384, 1998; [36] J. p. 9 A. Moreno and M. p. 9 Osorio, “Strict Lyapunov functions for the supertwisting algorithm,” IEEE Transactions on Automatic Control vol 57 no 4 pp 1035–1040, 2012; [38] Equation (38) follows from the LMI certificate of [36] at ∆ = ∆b i, ϕi ≤ ∆b i from Lemma 2 and Assumption 5 vol 7 no 1 pp 497–508; [40] Equation (40) is derived from the composite Lyapunov function analysis vol 3 no 3 pp 606–609. p. 9
pg1,Developing a demand-side management model that incorporates human decision-making and behavioral uncertainty into system operations, where the human acts as the leader in a leader-follower multi-agent system pg3,Designing a distributed observer for the human-in-theloop leader-follower multi-agent system with an unknown input, where the observer matching condition is not satisfied pg4,Performing stability analysis of the human-in-theloop leader-follower multi-agent system equipped with a distributed observer, operating under unknown input conditions pg4,The proposed leader-follower consensus control ensures system stability, while the human leader, acting as the facility manager, balances user thermal comfort and grid support pg10. p. 1
pg3,The system is modeled as a leader-follower MAS where cooperative buildings aim to track the state of the facility building, which receives DR signals from the utility pg3. p. 2
pg4,The core challenge involves incorporating a human as a leader, which introduces complexities due to variability and nonlinear responses, posing difficulties in achieving reliable control pg4. p. 0
pg4,To address the need for state estimation due to unmeasurable internal states like the derivative of the thermal comfort state, two primary observer approaches are considered: full-order distributed observers and interval observers pg4. p. 1
pg4,The paper notes that observer matching conditions, essential for standard linear unknown input observers, are not satisfied in this context because the unknown input is not directly reflected in the measured output pg4. p. 2
pg4,Consequently, sliding-mode observers are motivated as an alternative because they enforce a sliding manifold on the estimation error dynamics, compensating for unknown inputs without requiring the matching condition pg4. p. 4
pg3,The control objective is to coordinate participation in DR events while maintaining indoor temperatures within an acceptable range and ensuring consistent comfort levels across all participating buildings pg3. p. 3
pg4,A leader-follower consensus protocol is employed where the nominal control input uses a saturation function to respect physical actuator limits pg4. p. 3
pg3,The consensus error δi is defined by the interaction between neighbor states and the leader state pg3. p. 3
pg4,The observer design focuses on estimating consensus error using a distributed observer that processes local measurements and neighboring information to estimate this error pg4. p. 3
pg3,Simulation results validate the effectiveness of the proposed approach across three SCE-inspired DR scenarios: baseline DR, mid-course revision, and delayed commitment pg8. p. 6
pg4,Scenario A shows that the system successfully reduces demand by 300 kW while maintaining a new comfort state setpoint pg8. p. 6
pg4,Scenario B demonstrates robust consensus tracking control under abrupt changes in human control input by showing buildings quickly realign with the leader’s updated state after a revision pg8. p. 7
pg4,Scenario C illustrates the trade-off between grid support and comfort, showing that indoor temperatures can exceed the nominal comfort range near the end of an aggressive curtailment event pg8. p. 8
pg4,The simulation confirms that all transients remain bounded and decay rapidly, confirming stable observer performance under sudden operating changes pg8. p. 7
pg3,The proposed leader-follower consensus control ensures system stability while the human leader, acting as the facility manager, balances user thermal comfort and grid support pg10. p. 9
pg4,The results of the simulation study validate the effectiveness of the proposed approach, demonstrating the system’s ability to achieve consensus and maintain stability across various demand response scenarios pg10. p. 9
pg4,The results demonstrate that the proposed methodology increases the flexibility and adaptability of DR by seamlessly integrating human decision-making into advanced control systems, thereby improving overall performance and user satisfaction pg10. p. 9
pg4,This study paves the way for further exploration of human-in-the-loop control strategies in other complex, dynamic power system applications pg10. p. 9
pg4,The proposed methodology successfully achieves consensus and maintains stability across various demand response scenarios pg10. p. 9
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Human in-the-loop demand response management: The improved system can allow a human facility manager to dynamically balance conflicting objectives like occupant thermal comfort and grid support by setting participation levels for cooperative buildings, as described in the abstract and Section III A.
-
Distributed observer design for unknown input systems: The system can reliably estimate essential system states even when the
observer matching condition is not satisfied
by employing a sliding-mode observer tailored for leader-follower multiagent systems, as detailed in Section III C and D. -
Robust consensus control under human uncertainty: The improved leader-follower consensus protocol enables cooperative buildings to achieve consensus with the facility manager despite the facility manager's decisions being
unknown exogenous inputs that reflect complex trade-offs among thermal comfort, grid-service provision, and participation costs,
as stated in Section II A. -
Adaptive response to changing demand scenarios: The system can demonstrate robust tracking control under abrupt changes in human control input during mid-course revisions of demand response events, ensuring follower buildings
quickly realign with the leader’s updated state
by following the leader's trajectory, as shown in Simulation Scenario B.
Sources
Related papers
- Industrial Load Modeling and Optimization for Market-Based Interaction with Power Systems
- Exact and Approximate Convex Reformulation of Linear Stochastic Optimal Control with Chance Constraints
- Certified Set Convergence for Piecewise Affine Systems
- Symbolic Control for Stochastic Systems via Finite Parity Games
- Multiobjective optimization-based design and dispatch of islanded, hybrid microgrids for remote, off-grid communities in sub-Saharan Africa
- Physiology-informed layered sensing for intelligent human-exoskeleton interaction