Multi-agent Response Research Articles

Distributed Asynchronous Greedy Control of Large Networks of Thermostatically Controlled Loads for Electric Demand Response

This letter illustrates two multi-agent greedy demand-side response control schemes for networks of Thermostatically Controlled Loads. The objective is to provide simple but effective local control actions such that the overall power consumption tracks an aggregated desired profile. Compared with the existing literature the novelties are twofold. Since model-free, our schemes possess certain robustness features to the model deterioration and exogenous disturbances. Since greedy, they are of easy implementation also on cheap development boards which do not support optimization software, moreover because asynchronous do not require any network-wide synchronization event. Specifically, Algorithm 1 is very simple but it is applicable only on <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$K$ </tex-math></inline-formula> -regular communication topologies. Such prerequisite is then removed in Algorithm 2 by including within its instruction list a dynamic consensus protocol to estimate the mean network power consumption. Performance analysis and numerical simulations confirm the effectiveness of the schemes.

Resilience enhancement of multi-agent reinforcement learning-based demand response against adversarial attacks

Demand response improves grid security by adjusting the flexibility of consumers meanwhile maintaining their demand–supply balance in real-time. With the large-scale deployment of distributed digital communication technologies and advanced metering infrastructures, data-driven approaches such as multi-agent reinforcement learning (MARL) are being widely employed to solve demand response problems. Nevertheless, the massive interaction of data inside and outside the demand response management system may lead to severe threats from the perspective of cyber-attacks. The cyber security requirements of MARL-based demand response problems are less discussed in the existing studies. To this end, this paper proposes a robust adversarial multi-agent reinforcement learning framework for demand response (RAMARL-DR) with an enhanced resilience against adversarial attacks. In particular, the proposed RAMARL-DR first constructs an adversary agent that aims to cause the worst-case performance via formulating an adversarial attack; and then adopts periodic alternating robust adversarial training scenarios with the optimal adversary aiming to diminish the severe impacts induced by adversarial attacks. Case studies are conducted based on an OpenAI Gym environment CityLearn, which provides a standard evaluation platform of MARL algorithms for demand response problems. Empirical results indicate that the MARL-based demand response management system is vulnerable when the adversary agent occurs, and its performance can be significantly improved after periodic alternating robust adversarial training. It can be found that the adversary agent can result in a 41.43% higher metric value of Ramping than the no adversary case, whereas the proposed RAMARL-DR can significantly enhance the system resilience with an approximately 38.85% reduction in the ramping of net demand.

Linear Programming for Multi-Agent Demand Response

This research describes a real-time optimization model for multi-agent demand response (DR) from a Load Serving Entity (LSE) perspective. Three major categories of customers and five types of energy resources are considered simultaneously to achieve efficient DR decision making in highly stochastic future energy markets. Two infinite horizon stochastic optimization models are formulated; specifically, an LSE model and a dynamic pricing customer model. The objective of these models is to minimize long-term cost and discomfort penalty of the LSE and dynamic pricing customers. Because preferences of these two agents are different, they are inseparable and difficult to solve. A deterministic finite horizon linear program is solved as an approximation of the suggested stochastic model, and computational experiments are provided.

Adaptive multi-agent relief assessment and emergency response

The prediction of relief demand during disasters is an important operational task that should be performed before emergency response is initiated because such prediction enables the appropriate allocation and distribution of humanitarian supplies to affected zones. Current prediction methods appear to over or underestimate the calculation of relief demand—a problem that can be resolved by dispatching assessment vehicles to the affected zones of a region at the onset of a disaster. This study is aimed at facilitating the coordination of relief assessment and emergency response through the development of an adaptive multi-agent demand evaluation and demand-responsive model. Its main objective is to provide a model from which emergency response teams (ERTs) can obtain accurate information that will be used as basis by relief assessment teams (RATs) in effectively distributing humanitarian aid and conducting search and rescue operations. We propose a Markov decision process as a multi-agent assessment and response system, with reinforcement learning designed to ensure the integration of ERT and RAT operations. We use a coordination cluster system to coordinate ERT actors and use the proposed model to solve the problems occurring in a real-size network. Results show that the use of the model can improve emergency response operations and decrease death tolls.

Vanishing Price of Decentralization in Large Coordinative Nonconvex Optimization

We focus on nonconvex multi-agent optimization where a large number of participants collaboratively optimize some social cost function and their individual preferences. We focus on the semidecentralized multi-agent best response setting where a central planner attempts to coordinate the agents and each participant pursues self-interest. By studying the duality framework, we provide geometric and analytic characterizations of the duality gap and the price of decentralization. We prove that the nonconvex problem becomes increasingly convex as the problem scales up in dimension. Upon appropriate coordination, the price of decentralization asymptotically vanishes to zero as the number of participants grows. We develop a duality-based coordination procedure for the central planner to adapt the price vector and select a particular best response for each participant. The coordination algorithm is able to induce individual best responses to dynamically converge to an approximate global optimum, regardless of the ...

A multi-agent safety response model in the construction industry.

The construction industry is one of the sectors with the highest accident rates and the most serious accidents. A multi-agent safety response approach allows a useful diagnostic tool in order to understand factors affecting risk and accidents. The special features of the construction sector can influence the relationships among safety responses along the model of safety influences. The purpose of this paper is to test a model explaining risk and work-related accidents in the construction industry as a result of the safety responses of the organization, the supervisors, the co-workers and the worker. 374 construction employees belonging to 64 small Spanish construction companies working for two main companies participated in the study. Safety responses were measured using a 45-item Likert-type questionnaire. The structure of the measure was analyzed using factor analysis and the model of effects was tested using a structural equation model. Factor analysis clearly identifies the multi-agent safety dimensions hypothesized. The proposed safety response model of work-related accidents, involving construction specific results, showed a good fit. The multi-agent safety response approach to safety climate is a useful framework for the assessment of organizational and behavioral risks in construction.

Multi-agent cooperative intrusion response in mobile adhoc networks

The nature of adhoc networks makes them vulnerable to security attacks. Many security technologies such as intrusion prevention and intrusion detection are passive in response to intrusions in that their countermeasures are only to protect the networks, and there is no automated network-wide counteraction against detected intrusions. the architecture of cooperation intrusion response based multi-agent is propose. The architecture is composed of mobile agents. Monitor agent resides on every node and monitors its neighbor nodes. Decision agent collects information from monitor nodes and detects an intrusion by security policies. When an intruder is found in the architecture, the block agents will get to the neighbor nodes of the intruder and form the mobile firewall to isolate the intruder. In the end, we evaluate it by simulation.

ICONE15-10572 Development of Emergency Response System for Disaster Management

To response large scale disaster, for example hurricane, earth-quake, or terrorist attack, it is usual case that multi-organizations, such as government (central or local), commercial entities, media, and public are involved. How to make these inter- or intra dependent organizations cooperate timely and accurately during crisis is the primary concerns. In this study, we model multi-organizations cognitive process, which introduced organizational factors, individual factors and focused on how to effectively incorporate human teams into a socio-technological system. Based on it we develop a multi-agent emergency response system. The test simulation is carried out using scenarios extracted from real emergency drills.