Another example of multi-agent systems is trading. To illustrate this,
we recently built a demonstration for simulating an electric energy
trading market during an extreme weather event within New York
State. In this project, we used AI agents that acted on behalf of each
of the participants in the market.
For example, there were agents that bid into the market on behalf of
generators, with the goal of maximizing profits. Other agents
represented the buyers, like a local utility company that wants to buy
electricity at the lowest possibly prices without disrupting service to
their residential and commercial customers.
A third type of agent represented the interests of the market maker to match buy/sell bids that best ensure grid reliability. In this case, it was the organization charged with the operation of the New York State transmission grid (NYISO).
By running these agents in simulation mode, we were able to simulate the behavior of the market and enable "what-if" analyses.
Human Oriented Software
These examples of AI Agents illustrate the evolution of software solutions
towards a human orientation that deeply understands and reasons about
underlying systems of interest and their complex problems.
We at SmartCloud believe in this evolution and the need for AI Agents in
the face of growing computing ubiquity, interconnection, and autonomy.
We're committed to being at the forefront of this AI technology and bringing
its full value to the business world.
Even Hulk and Hawkeye would have to agree with that kind of
Why Multi-agent Systems?
There are several trends in our society which are driving the need for
AI agents and multi-agent systems. As a response to these trends,
we at SmartCloud have adopted AI agent technology and multi-agent
systems within the core of our business solutions.
The first of these trends is ubiquity in computing power. You see it
everywhere – in your home (phones, thermostats, Internet, etc.),
Sports Agents, Secret Agents and
There are many kinds of agents.
In the 1996 movie Jerry Maguire, the title character is played by
Tom Cruise. He is a sports agent for the Cardinal's wide receiver,
Rod Tidwell, played by Cuba Gooding, Jr. The best way for his
agent to represent him, argues Tidwell, is to "Show me the money."
M1 gives a new assignment to the
British Intelligence Agency's "007,"
also known as "Bond – James Bond."
I In the Marvel Cinematic Universe,
there is an organization called
Strategic Homeland Intervention,
Enforcement and Logistics
Division (S.H.I.E.L.D.). It is a fictional
peacekeeping and spy agency dealing
with numerous superheroes and
supernatural phenomena. The director of S.H.I.E.L.D. needed some
super help, so he called on a powerful group of agents called the
Avengers. The group included some admirable characters like
Tony Stark (Iron Man), the Hulk, Hawkeye, Thor, Captain America
and Black Widow. Each one had a unique superpower to offer.
So what is an agent? It is something that represents someone or
something else. An agent acts on behalf of its owner, sometimes
simply just following its owner's instructions but often with a good
amount of independence to save the owner time and get better results.
In the world of "Artificial Intelligence" software you also find agents –
AI agents (also referred to as intelligent agents), which are emerging
as a powerful programming technique for creating flexible and
scalable solutions. These kinds of agents have goals, strategies for
achieving these goals, and tasks they undertake to carry out
Today you'll find AI agents acting on behalf of humans within cars
and planes, hospitals and labs, computers and robots, to name a few.
They are also used in industry to control machine operations, and in space to remotely control spacecraft and probes.
More recently, additional AI agents are being added to the mix to create unified systems of agents that can do much more than single agents alone. By emulating how humans collectively make decisions to achieve goals, like they do as part of a supply chain, production operation, electric distribution grid, or trading market, multiple AI agent systems can greatly improve business results.
These kinds of multi-agent systems, like their human counterparts, can organize themselves in hierarchies, including with supervisory agents, to best reach overall objectives for a business. For instance low-level agents acting on behalf of factory workstations might focus on achieving maximum flow of their station's production while their supervisory agents are focused on the decisions that lead to an optimal balance of factory-wide production and costs.
This type of agent organization also has the advantage of distributing computing to enable decision making within big data environments. Lower-level agents do what might be described as "little compute" on devices and supervisory agents do a "big compute" on a central server, such as one in an Internet cloud.
September 1, 2015
your car (cruise control, GPS, engine metrics like fuel consumption, etc.) and your work (factories, offices, equipment, networks, etc.). AI Agents are naturals for getting the most out of all this computing.
There is more and more interconnection, with the interchange of data becoming cheaper and cheaper. Mobile phone companies compete for the most attractive data plans, and we all benefit. Businesses connect more and more things to improve productivity. But with growing interconnection, comes growing complexity of tasks and decisions, and the need for more and more intelligent solutions that work in real time. AI agent system solutions thrive in managing these kinds of complexities.
Another driving force is autonomy of computing. A smart home thermostat
(like a Nest), an airplane autopilot system, or self-driving vehicles are
good examples of delegating responsibility to computers – with little, if any,
human intervention. By doing this kind of delegation wherever possible,
we will begin to experience transformational results. AI Agents are
designed to work autonomously as they make decisions and undertake
tasks in ways that are not possible using traditional software programming.
Multi-agent systems at SmartCloud
We use AI agents in different ways. Our agents are speedy – they all work in real time, down to subsecond levels, as they make decisions and take actions.
For example, one of our solutions is helping operators with the North American Electric Reliability Corporation (NERC) watch over the U.S. bulk power system to proactively detect and respond to problems in real time. That bulk power system is a series of high voltage transmission power lines and generators across the U.S. and Canada, servicing over 330 million people.
There are AI agents for over 30,000 geographically dispersed data points that are processing the data for accuracy and timeliness by intelligently taking into account issues of latencies, delays, corruptions, and duplications.