Comment Intelligent agents (Score 3, Informative) 67
Some of the below has been hacked from my thesis, so please excuse the lack of references (the citation have been deleted).
When AI research began to consider the possibilities of distributed applications, the field of distributed AI (DAI) emerged. Within this field, there are three general areas: Distributed Problem Solving (DPS), Multi-Agent Systems (MAS) and Parallel AI. Agents in DPS are low-grain, often sharing common resources. Agents in a MAS are large-grain, having autonomy and heterogeneity and are typically based the psychology-based Belief-Desire-Intention (BDI) model. Although the BDI model is considered a robust and flexible way of describing the internal state of intelligent agents, it is complicated and difficult to implement. This gap between BDI theory and practice means that the model has to be extended for the development of practical goal directed agents. Furthermore, there is no general architecture in MAS research. Consequently, MAS are complex to construct and are usually built for a specific purpose. They are heterogeneous, that is, an agent from one MAS is inherently incompatible with another MAS. The field of Cooperating Knowledge-Based Systems (CKBS) provides a general architecture for the development of real-world systems (for example, where database are heavily used), and inherit the benefits of DAI: modularity, parallelism and reliability (due to redundancy).
There have been various definitions of an agent and agent-based systems. Norvig defined agents as intelligent entities that can be "viewed as perceiving its environment through sensors and acting upon its environment through effectors" Others view agents from a more practical perspective, as software engineering solutions to complex problems. The Oxford Dictionary of Computing defined an agent as 'an autonomous system that receives information from it's environment, processes it, and performs actions on that environment. Agents may have different degrees of intelligence or rationality, and may be software, hardware, or both.' The point has been made that, under the banner of 'agents', the research is truely heterogeneous.
Four important characteristics of a MAS were identified by Sycara and these may be applied to the pragmatic CKBS perspective:
- Each agent has incomplete information or capabilities for solving the problem and, thus, has a limited viewpoint.
- There is no system global control.
- Data are decentralised.
- Computation is asynchronous.
The InfoSleuth project stated that the use of agents provides a 'high degree of decentralisation of capabilities which is the key to system scalability and extensibility.' This is due to fewer resource limitations, fewer communication bottlenecks and the absence of a single point of failure. Furthermore, modules may be added easily, so the system is more scalable and there may be more than one agents able to perform a task; this redundancy begets reliability.
Agents have been used in a wide variety of scenarios: manufacturing, telecommunications, air traffic control, information gathering. There are increasingly more companies involved in non-research agent activities, for example http://www.lostwax.com/.
When AI research began to consider the possibilities of distributed applications, the field of distributed AI (DAI) emerged. Within this field, there are three general areas: Distributed Problem Solving (DPS), Multi-Agent Systems (MAS) and Parallel AI. Agents in DPS are low-grain, often sharing common resources. Agents in a MAS are large-grain, having autonomy and heterogeneity and are typically based the psychology-based Belief-Desire-Intention (BDI) model. Although the BDI model is considered a robust and flexible way of describing the internal state of intelligent agents, it is complicated and difficult to implement. This gap between BDI theory and practice means that the model has to be extended for the development of practical goal directed agents. Furthermore, there is no general architecture in MAS research. Consequently, MAS are complex to construct and are usually built for a specific purpose. They are heterogeneous, that is, an agent from one MAS is inherently incompatible with another MAS. The field of Cooperating Knowledge-Based Systems (CKBS) provides a general architecture for the development of real-world systems (for example, where database are heavily used), and inherit the benefits of DAI: modularity, parallelism and reliability (due to redundancy).
There have been various definitions of an agent and agent-based systems. Norvig defined agents as intelligent entities that can be "viewed as perceiving its environment through sensors and acting upon its environment through effectors" Others view agents from a more practical perspective, as software engineering solutions to complex problems. The Oxford Dictionary of Computing defined an agent as 'an autonomous system that receives information from it's environment, processes it, and performs actions on that environment. Agents may have different degrees of intelligence or rationality, and may be software, hardware, or both.' The point has been made that, under the banner of 'agents', the research is truely heterogeneous.
Four important characteristics of a MAS were identified by Sycara and these may be applied to the pragmatic CKBS perspective:
- Each agent has incomplete information or capabilities for solving the problem and, thus, has a limited viewpoint.
- There is no system global control.
- Data are decentralised.
- Computation is asynchronous.
The InfoSleuth project stated that the use of agents provides a 'high degree of decentralisation of capabilities which is the key to system scalability and extensibility.' This is due to fewer resource limitations, fewer communication bottlenecks and the absence of a single point of failure. Furthermore, modules may be added easily, so the system is more scalable and there may be more than one agents able to perform a task; this redundancy begets reliability.
Agents have been used in a wide variety of scenarios: manufacturing, telecommunications, air traffic control, information gathering. There are increasingly more companies involved in non-research agent activities, for example http://www.lostwax.com/.
