The Usage of Intelligent Agents in IoT devices
Using assisted learning within user-centered IoT devices represents a step into the future in terms of how communication is performed. The main benefit intelligent agents bring to the IoT table is the creation of a system based on users' needs. As a result, the different gadgets can learn the user's habits, leading to more efficient usage of that gadget.
IoT devices have developed in an intensive rhythm, and interconnected devices have started to offer data about the things they work on. Nevertheless, data is not valuable unless it is transformed into information which can later be used. This refers back to the Big Data field where raw data is processed to obtain valuable information. Yet, it takes time and research to be able to obtain knowledge about what was done and what should be done.
AI has boomed because of the processing power which has become bigger and bigger, while the size of calculus systems has shrunk to the level of an USB drive. Evolution also brought about intelligent agents, which became more and more intuitive due to different learning techniques.
Stuart Jonathan Russel states that: "An agent is an entity which perceives the surrounding environment and acts on it"[1].
Agents are autonomous and adaptive software products which can be used to perceive the environment they come from. With the help of sensors, based on their input knowledge, and using assisted learning, they accumulate positive and negative knowledge about the environment they act in. Based on the information they later gather or observe, they can take decisions and enable these decisions via effectors.
Assisted learning is the process by which agents are helped to take decisions based on the feedback received from the users or the system they come from, in the different, unknown situations which come about. After the agent develops a knowledge base, when one of the learnt situations occurs again, the agent will know what to do. Moreover, the learning method and the decisions the agent takes are ever changing and improving.
The diagram below defines how an IoT system with an integrated intelligent agent should look like. The diagram also details its functionality, associated to the learning method and the accumulation of new knowledge.
When the IoT system is used by the user in an environment, the sensors in the device collect data from the outside. The data is submitted to an agent for knowledge analysis and extraction, so that the agent knows how to handle the environment next time.
Figure 1.
Learning Element is the process responsible for improving the agent's learning method so that the Performance Element is as efficient as possible. The "Critic Element" is essential for the agent to know it took the right decision or not. This component evaluates how good the decisions taken by the agent really are. Feedback is then sent to the Learning Element. The Learning Element is directly connected to the Problem Generator which studies the possibilities that have not yet occurred during the functioning or learning processes of the agent. The Critic and Problem Generator processes can be considered essential elements in the agent's learning process.
Coming back to the Learning Element, it sends the information it learned onto the Performance Element the component which decides which actions are good and what needs to be sent onto the Effector. Effectors apply or implement the decision actions taken by the agent, these being destined for the end user. Given the taken decisions, the "Performance Element" decides whether the actions were good or not, and it sends feedback onto the "Learning Element" which recreates the knowledge base.
Until now, the concept was implemented only by IT giants, for the purpose of developing new technologies. The examples from real life, which use this concept, are:
The autonomous car designed by Google, which is driverless, whose sensors collect data from the surrounding environment, and which, based on the data, can decide what course of action needs to be taken, by handling unknown situations.
Siri, Cortana, Bixby, based on the data retrieved from the user and the environment, assist the user in the decision-making process.
In conclusion, in terms of decision-making by devices, assisted learning and intelligent agent integration in IoT devices is beneficial, as it offers the user great comfort. The devices learn the habits of people and satisfy their needs. Therefore, we can say that this technology represents a major step into the future and towards intelligent devices which know what functions they must fulfill, by adapting to the users' profile.
Artificial Intelligence A MODERN APROACH, Third Edition, Stuart Jonathan Russel & Peter Norvig, Editura: Pearson
by Andrei Oneț
by Paul Bodean , Eugen Meltis
by Dan Sabadis
