2. The Requirements of an Integrated Environment for Artificial Intelligence

An agent is an entity that perceives characteristics of its environment and acts upon that environment. An intelligent agent acts upon its environment in ways humans consider appropriate to the characteristics that the agent perceives. It uses its sensors - means of perceiving its surroundings - to collect information that it then uses to make intelligent decisions. The agent then acts upon its surroundings through its effectors [1]. It is in this decision making, or mapping of input from sensors to output actions through the effectors, that the intelligence of the agent lies. This mechanism for agent intelligence, however, is not the focus of this research; there are many different approaches and techniques for making an agent intelligent, which encompass several major paradigms and philosophies. It is also not the place of this research to make a judgment as to which paradigm is most appropriate for SHELLEY, but rather to design an environment that can facilitate different approaches to building an intelligent agent so that future students can explore the many options without feeling confined or restrained to one predefined paradigm.

The distinction can be made between pure software agents, whose world consists entirely of entities internal to the computer upon which the agent resides (soft agents), and agents whose world extends beyond the confines of a computer and encompasses the physical parameters of its surroundings [1]. In the former, the sensors and effectors are much more easily implemented, while the latter requires special hardware that introduces all the complications already discussed (see Motivations).

SHELLEY is a robotic entity - her environment is the physical world. Therefore, while some tasks may be handled sufficiently through soft agents, others must be addressed by agents who make use of SHELLEY's special hardware peripherals in order to interact with the 'real world' [2]. In light of the aforementioned difficulties associated with interfacing to these peripherals, we require some mechanism through which agents can easily access specialized external hardware in order to accomplish their tasks. Thus, SHELLEY necessitates an integrated environment that can provide an effective and flexible system for integrating both existing and future peripherals such that these devices can be shared and adequately managed. Additionally, this integrated environment must provide a simple method of programming with and using these peripherals. Ideally, this can exist in the form of function calls which can be made directly from within researchers' program code; however, the function calls must be structured such that they can easily accommodate new and different types of devices, as well as be used from any one of numerous programming languages.



Next     Previous Contents