Scribd
Upload
19 views11 pages

Multi-Agents For CPS Based Smart Building Control: Saurav Bhattarai Advisor: Prof. Marcelo Simões

The document proposes a multi-agent system approach for smart building control. It involves three agents - Electrical, Heating/Cooling, and Comfort that work together towards a global system goal. Each agent is connected to different physical systems in the building to perform actions and sense the environment. The agents communicate with each other to coordinate their actions while controlling costs through utility functions and agent behaviors in order to reduce redundancy between agents. Future work includes implementing a communication scheme between agents and building the full algorithm for calculating optimal policies.

Uploaded by

saurav_mobile
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
19 views11 pages

Multi-Agents For CPS Based Smart Building Control: Saurav Bhattarai Advisor: Prof. Marcelo Simões

The document proposes a multi-agent system approach for smart building control. It involves three agents - Electrical, Heating/Cooling, and Comfort that work together towards a global system goal. Each agent is connected to different physical systems in the building to perform actions and sense the environment. The agents communicate with each other to coordinate their actions while controlling costs through utility functions and agent behaviors in order to reduce redundancy between agents. Future work includes implementing a communication scheme between agents and building the full algorithm for calculating optimal policies.

Uploaded by

saurav_mobile
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 11

Multi-Agents for CPS Based Smart Building Control

Saurav Bhattarai Advisor: Prof. Marcelo Simes

Smart Building Characteristics


Buildings are complex structures
Electrical System Heating/Cooling System Sensing

Communication
Constant communication between different systems

Large amounts of data


Required for operation Produced Processing takes a long time

Legacy Systems
Existing buildings do not have all the infrastructure for being smart

Multi-Agent Systems Approach


Agents
Intelligent Autonomous Work towards goals Pro-active Reactive

Advantages
Distributed computing resources Failures are not system-wide Modular Parallelism

Decentralized Control

Overall Control Structure

Proposed Structure
Multi-agent approach Three agents
Electrical Heating/Cooling Comfort

Controlling mechanism
Cost functions Agent Behavior Reduces redundancy between agents

Multi-Agent Structure

The proposed MAS structure includes 3 agents: Electrical, Heating/Cooling, and Comfort. All the agents have communication channels between them. This allows for the agents to work together to reach the global system goal. Agents are connected to the different physical systems, allowing them to perform actions on and sense the building environment

Heating Agent Input/Output

Electric Agent Input/Output

Error tolerance = 5% 50% reduction of expected utility per state change (=0.5) o Utility values settled in 6 iterations 30% reduction of expected utility per state change(=0.7) o Utility values settled in 12 iterations o More flexibility to move between states to maximize utility value

A scenario involving the comfort agent was selected o Temperature control Transition values (probabilities) assigned for actions Utility values of each state calculated using Value Iteration algorithm to solve Bellman update equation.
u t 1 ( s ) r ( s ) max T ( s, a, s ' )u t ( s ' )
a a'

Future Work
Use transition values obtained from studying a physical system to calculate realistic utility values Implement a communication scheme between multiple agents Define states using Fuzzy Logic Build full algorithm for calculating optimal policies for all agents

You might also like