Den här sidan är utskriven från Högskolan i Halmstads webbplats (www.hh.se). Texten uppdaterades senast den 2017-09-13. Besök webbplatsen om du vill vara säker på att läsa den senaste versionen.
The scientific focus for the Center for Applied Intelligent System Research (CAISR) is “aware” intelligent systems: human aware, situation aware, and self-aware. Aware intelligent system will become even more important in the “ubiquitous future with data, sensors and embedded computers everywhere Also, aware intelligent systems will be ubiquitous in applications in health care and vehicles.
The overall common research theme for CAISR is aware intelligent systems. The goal with artificial intelligence (AI) research and development is to construct systems that behave intelligently and transform the way we live and work (Machine Learning is included in AI). However, AI solutions tend to be brittle; they break when the reality deviates from what was anticipated by their designers. The standard paradigm in AI is supervised learning, i.e. assume that human experts are available to define the task that the system should perform and teach it to replicate this and as best as possible generalize from the examples. Obviously, this assumption never holds in real life. It is impossible to have human experts that can label a huge corpus of data. Furthermore, things change. This limits the applicability of supervised AI.
Our aim is to go beyond this and approach the construction of AI systems that work in a variety of situations. Real life is complex but real life is where we want our systems to operate. In order to do so, the systems must become more “aware” and able to learn on their own. If a system is to be able to adapt to changing circumstances, it must be aware of the circumstances. Furthermore, new data is generated continuously and we need systems that can, in an autonomous fashion, deal with this deluge of data.
Awareness refers to being higher in the so-called knowledge pyramid, having moved up from experiencing sensor sensations to the levels of integrating them into knowledge and understanding. Awareness also refers to self-organizing, the ability to learn and act autonomously, and to an awareness control loop with monitoring → recognition → assessment → learning → monitoring, and so on, endlessly, building up knowledge, becoming aware.
A good and highly relevant application example is the assisted living environment for elderly (ambient assisted living). In its simplest form is it just collecting sensor data and transferring them to a human for interpretation and dialogue, e.g. a remote controlled mobile robot with a smartphone or an emergency alarm system (with a button to push). In a slightly more advanced form it is equipped with sensors (motion, door, heat, smoke, etc.) and can issue an alarm based on predefined rules for the sensor readings. In an even more advanced form can the system learn the signal patterns from the sensors when the inhabitant is at home feeling well and alarm when the observations deviate from this. In this latter example is the system situation aware and has learned this in a self-organized way, unloading tedious monitoring work from human personnel. With time can observed abnormality be associated with particular events (e.g. visitors) and a knowledge base be built up. An even more advanced version also has cameras and can detect where humans are, can react to human signals and interact with the human (e.g. checking that everything is ok, and issuing simple alarms). Here the system exhibits a human-aware ability. Going even further up the knowledge triangle can mean that the system is able to interact with the inhabitant and provide cognitive training and “keeping company”.
If a system is to be able to adapt to changing circumstances, it must be aware of the circumstances.