ݺߣ

ݺߣShare a Scribd company logo

EcoHome

•Download as DOCX, PDF•
•
D
devsunit

The document discusses using artificial intelligence techniques to control devices in a smart home. It describes how voice recognition, image recognition, pattern recognition, and expert systems could be used by an intelligent agent to monitor the home, learn user preferences, and automate tasks like temperature control. The agent would collect data from sensors to recognize patterns in user behavior and optimize home operations. Networks like WiFi would connect all devices to allow centralized control and automation through the intelligent agent.

1 of 2
Download to read offline
EcoHome Control Things Using AI Techniques Murtaza Ahmed Kazi Student at SZABIST Bs in Computing Karachi, Sind, Pakistan kazimurtaza@gmail.com Abstract— the concept of Smart Homes using AI plays an important role in the planning of future housing-based models of care. More and more research groups are working in this area, Such as MIT, Siemens, Cisco, IBM, Xerox, Microsoft and Etc. In these groups, nearly 20 home labs have been set. In These home labs, more than 30 appliances are used, over 5Network protocols are produced and more than 3 AI Techniques have been used. Keywords— Artificial Intelligence; Voice Recognition; Image & Pattern Recognition; Expert System; Android; Arduino I. INTRODUCTION Smart house; devices integrated with an intelligent agent. Letting the agent control our environment would not only bring efficiency but comfort, like security management, and electrical appliance control and other all sorts of controls one could think of. Agent could observe the patterns and would perform actions on prediction if given the platform. Many believe that we should use voice recognition to communicate with the agent, control devices operation with your voice such as, “living-room light1 on”, agent could easily split the string and classify it as living-room being the location and light1 as device and on/off as action to perform on it. This means agent would require understand discrete words, but that is not the difficult part as explained above, the part which could cause problem is teaching all possible input, but keeping in mind since our model world is a typical home, we could easily classify it for an agent, how many rooms and devices there are and what are their identifiers, such as living room, kitchen, master bedroom, second bedroom and others. II. VOICE RECOGNITION[1] The part where voice is converted to text could become a challenge. We could easily use tools available to us, rather than reinventing the wheel, voice tool which could be used is already available in cheapest of android phones, “goggle voice” or “goggle now” - it is open source and could be used in an android app with two lines of code, so we would not need to make a language processing module. Voice could eliminate all unnatural form of communication. And voice recognition is not evolved enough to sort out speech patterns, which could be a limiting factor, But luckily for us, Android phones already comes with voice recognition and processing framework, which could be used to take input from user and send to a agent which would be connected to all the devices, where as a agent could be a computer or a micro-controller such as Arduino or Netdino(with SD- Card, could hold knowledge base which has all the knowledge of all devices in the house and its operation once it gets the input from owners phone it could then send out the preferred action to the device. III. EXPERT SYSTEMS[2] one could say this could be perfect environment for a expert system, it could manage the home in comfortable way it could prove to be quiet efficient to perform some of the goals like managing temperature of the house while conserving energy, by using temperature preferred by the owner, present temperature, and humidity data or conserving energy using motion sensors to check if no one is around to turn of the room. System could easily keep this data and adapt it by learning process, sensors such as temperature and humidity and motion could be used to easily provide such information to the system. This system could have operation such as; prediction of owner’s preferences, learning efficient ways of doing operation like for example switching on devices in morning before the owner has set the alarm for, this way things could be ready for him when he does wake up. All is only possible when everything is connected centrally and all information could be accessed by the agent like alarm clock, motion, temperature, time, lights, and owner’s preferences and other. Expert system and combination of rules, knowledge base and an inference mechanism for applying the rules, knowledge base would be made of knowledge of house such as – motion sensing, devices and its types, other systems, thermal details and previous consumptions of things and its inhabitants their names relationships, age, priority habits and preferences. And system could not generate this knowledge base by just observing actions, we need to classify what should already be taught to the agent and what things, the agent can learn on its own. IV. PATTERN RECOGNITION[3] Once decided what things an agent should learn from, now let’s look how an agent would observe, Pattern recognition; supplying of data to system, one could use this to recognize the habits of the owner and, using pattern recognition to learn how many people are in house to adjust the temperature could be a example and gathering information from sensors and making patterns for owner, like when he usually wakes up, so agent could bring the temperature to the preferred point or turning on the water heater or if owner has suddenly fallen to ground and is not answering back to the agent perhaps the owner needs some sort of help and call emergency services, while no such sensors are available we could use other sensors in combination to determine things, such as camera, proximity sensor, motion sensor, light sensor, temperature sensors and global position system , dirty location provided by networks and wifi with goggle maps using overlays to find accurate location to a specified room of the owner. For example a house could be in one of two states occupied or unoccupied, by knowing this the system could modify its operation like if the home is unoccupied it would switch of
all devices and turn on the security system. Integrating a camera, could open up a whole channel of information for agent to compute although having maximum data could make a very efficient agent, but will also over crowd it and slow down its computation, suppose we have no hardware limitations for now and we integrate camera with agent, which could possibly recognize faces, habits and position of its inhabitants quiet efficiently, But processing image is quiet expensive computation wise and agent has to be at least a desktop machine. One example of Image processing; could be that camera checks the person who just entered the room and scans its faces and finds no one with such face in its knowledge base could call police and alert owners or others and starts playing loud music and tells the person to evacuate and its picture has been recorded and sent to local police. V. NETWORK[4] And at last we only require one thing which will be able to connect all these devices together much like “bus” in motherboards of our computers, creating smart house network, option available are; Powerline (XlO, EIB Powerline etc); Busline: EIB, Cebus, Lonwork, Batibus EHS etc) and Radio Frequency (RF) (e.g. Bluetooth, Wifi, Zigbi and most major smart home manufacturers), choosing one totally depends on your situation powerlines and buslines are not appropriate updates for a house which is already made that situation could use radio frequency more efficiently preferably wifi as others have short range and capacity or zigbi which has the ability to form a mesh network. VI. EXAMPLES[5] Some of the examples could be; Security Management System controlled by AI, keeping track of its owners and protecting them in case of all sorts of emergencies, and being a personal assistant of all its owners, keeping track of their calendar and schedules and reminding tasks which are given by owner for another owner, and controlling appliances by itself or by owner request, using voice as communication medium between agent and owner, keeping track of electricity, gas and water usage and forwarding it to a local server for detailed charts and statistical computation and representations, and parental control – letting parents know where their kids are what are they doing. And some smart devices could be; a pillow with in-built speakers, which could be used for music and reading book while owner falls asleep, and keep check sensors for data such as respiration, temperature, pulse and if something is wrong could warn emergency services or others living in the house. And refrigerator which keeps data about all the items in side and keep track of expiration data and could place order when something is finished could use camera with image processing to keep track of items. VII. CONCLUSION Voice Recognition, Image recognition, pattern recognition, Expert system all require one essential component, Learning, observing patterns and changing knowledge base and rules to meet owners feedback/preferences and efficient standards. Learning could be done in two ways one is that a user requests something to be canceled or second using of patterns recognition to learn something by observing user pattern. REFERENCES [1] Enhancingresidential automationsystems with artificial intelligence,T. Abinger W. KastnerG. Luber G. Neugschwandtner, Automation Systems Group, Vienna Universityof Technology,Siemens AG, NX Scientific Conference 2008,November2008. [2] Solvinghome automationproblems usingartifical intelliengence, Grayson Evans - Directorof Engineering, MTI Inc.,Portland, OR. pp. 396, August 1991. [3] Solvinghome automationproblems usingartifical intelliengence, Grayson Evans - Directorof Engineering, MTI Inc.,Portland, OR. pp. 398, August 1991. [4] Smart home research, Li Jiang, Da-YouLiu, Bo Yang, College of Computer Science andTechnology, Jilin University,pp 659, August 2004. [5] Smart home research, Li Jiang, Da-YouLiu, Bo Yang, College of Computer Science andTechnology, Jilin University,pp 660, August 2004.

More Related Content

EcoHome

  • 1. EcoHome Control Things Using AI Techniques Murtaza Ahmed Kazi Student at SZABIST Bs in Computing Karachi, Sind, Pakistan kazimurtaza@gmail.com Abstract— the concept of Smart Homes using AI plays an important role in the planning of future housing-based models of care. More and more research groups are working in this area, Such as MIT, Siemens, Cisco, IBM, Xerox, Microsoft and Etc. In these groups, nearly 20 home labs have been set. In These home labs, more than 30 appliances are used, over 5Network protocols are produced and more than 3 AI Techniques have been used. Keywords— Artificial Intelligence; Voice Recognition; Image & Pattern Recognition; Expert System; Android; Arduino I. INTRODUCTION Smart house; devices integrated with an intelligent agent. Letting the agent control our environment would not only bring efficiency but comfort, like security management, and electrical appliance control and other all sorts of controls one could think of. Agent could observe the patterns and would perform actions on prediction if given the platform. Many believe that we should use voice recognition to communicate with the agent, control devices operation with your voice such as, “living-room light1 on”, agent could easily split the string and classify it as living-room being the location and light1 as device and on/off as action to perform on it. This means agent would require understand discrete words, but that is not the difficult part as explained above, the part which could cause problem is teaching all possible input, but keeping in mind since our model world is a typical home, we could easily classify it for an agent, how many rooms and devices there are and what are their identifiers, such as living room, kitchen, master bedroom, second bedroom and others. II. VOICE RECOGNITION[1] The part where voice is converted to text could become a challenge. We could easily use tools available to us, rather than reinventing the wheel, voice tool which could be used is already available in cheapest of android phones, “goggle voice” or “goggle now” - it is open source and could be used in an android app with two lines of code, so we would not need to make a language processing module. Voice could eliminate all unnatural form of communication. And voice recognition is not evolved enough to sort out speech patterns, which could be a limiting factor, But luckily for us, Android phones already comes with voice recognition and processing framework, which could be used to take input from user and send to a agent which would be connected to all the devices, where as a agent could be a computer or a micro-controller such as Arduino or Netdino(with SD- Card, could hold knowledge base which has all the knowledge of all devices in the house and its operation once it gets the input from owners phone it could then send out the preferred action to the device. III. EXPERT SYSTEMS[2] one could say this could be perfect environment for a expert system, it could manage the home in comfortable way it could prove to be quiet efficient to perform some of the goals like managing temperature of the house while conserving energy, by using temperature preferred by the owner, present temperature, and humidity data or conserving energy using motion sensors to check if no one is around to turn of the room. System could easily keep this data and adapt it by learning process, sensors such as temperature and humidity and motion could be used to easily provide such information to the system. This system could have operation such as; prediction of owner’s preferences, learning efficient ways of doing operation like for example switching on devices in morning before the owner has set the alarm for, this way things could be ready for him when he does wake up. All is only possible when everything is connected centrally and all information could be accessed by the agent like alarm clock, motion, temperature, time, lights, and owner’s preferences and other. Expert system and combination of rules, knowledge base and an inference mechanism for applying the rules, knowledge base would be made of knowledge of house such as – motion sensing, devices and its types, other systems, thermal details and previous consumptions of things and its inhabitants their names relationships, age, priority habits and preferences. And system could not generate this knowledge base by just observing actions, we need to classify what should already be taught to the agent and what things, the agent can learn on its own. IV. PATTERN RECOGNITION[3] Once decided what things an agent should learn from, now let’s look how an agent would observe, Pattern recognition; supplying of data to system, one could use this to recognize the habits of the owner and, using pattern recognition to learn how many people are in house to adjust the temperature could be a example and gathering information from sensors and making patterns for owner, like when he usually wakes up, so agent could bring the temperature to the preferred point or turning on the water heater or if owner has suddenly fallen to ground and is not answering back to the agent perhaps the owner needs some sort of help and call emergency services, while no such sensors are available we could use other sensors in combination to determine things, such as camera, proximity sensor, motion sensor, light sensor, temperature sensors and global position system , dirty location provided by networks and wifi with goggle maps using overlays to find accurate location to a specified room of the owner. For example a house could be in one of two states occupied or unoccupied, by knowing this the system could modify its operation like if the home is unoccupied it would switch of
  • 2. all devices and turn on the security system. Integrating a camera, could open up a whole channel of information for agent to compute although having maximum data could make a very efficient agent, but will also over crowd it and slow down its computation, suppose we have no hardware limitations for now and we integrate camera with agent, which could possibly recognize faces, habits and position of its inhabitants quiet efficiently, But processing image is quiet expensive computation wise and agent has to be at least a desktop machine. One example of Image processing; could be that camera checks the person who just entered the room and scans its faces and finds no one with such face in its knowledge base could call police and alert owners or others and starts playing loud music and tells the person to evacuate and its picture has been recorded and sent to local police. V. NETWORK[4] And at last we only require one thing which will be able to connect all these devices together much like “bus” in motherboards of our computers, creating smart house network, option available are; Powerline (XlO, EIB Powerline etc); Busline: EIB, Cebus, Lonwork, Batibus EHS etc) and Radio Frequency (RF) (e.g. Bluetooth, Wifi, Zigbi and most major smart home manufacturers), choosing one totally depends on your situation powerlines and buslines are not appropriate updates for a house which is already made that situation could use radio frequency more efficiently preferably wifi as others have short range and capacity or zigbi which has the ability to form a mesh network. VI. EXAMPLES[5] Some of the examples could be; Security Management System controlled by AI, keeping track of its owners and protecting them in case of all sorts of emergencies, and being a personal assistant of all its owners, keeping track of their calendar and schedules and reminding tasks which are given by owner for another owner, and controlling appliances by itself or by owner request, using voice as communication medium between agent and owner, keeping track of electricity, gas and water usage and forwarding it to a local server for detailed charts and statistical computation and representations, and parental control – letting parents know where their kids are what are they doing. And some smart devices could be; a pillow with in-built speakers, which could be used for music and reading book while owner falls asleep, and keep check sensors for data such as respiration, temperature, pulse and if something is wrong could warn emergency services or others living in the house. And refrigerator which keeps data about all the items in side and keep track of expiration data and could place order when something is finished could use camera with image processing to keep track of items. VII. CONCLUSION Voice Recognition, Image recognition, pattern recognition, Expert system all require one essential component, Learning, observing patterns and changing knowledge base and rules to meet owners feedback/preferences and efficient standards. Learning could be done in two ways one is that a user requests something to be canceled or second using of patterns recognition to learn something by observing user pattern. REFERENCES [1] Enhancingresidential automationsystems with artificial intelligence,T. Abinger W. KastnerG. Luber G. Neugschwandtner, Automation Systems Group, Vienna Universityof Technology,Siemens AG, NX Scientific Conference 2008,November2008. [2] Solvinghome automationproblems usingartifical intelliengence, Grayson Evans - Directorof Engineering, MTI Inc.,Portland, OR. pp. 396, August 1991. [3] Solvinghome automationproblems usingartifical intelliengence, Grayson Evans - Directorof Engineering, MTI Inc.,Portland, OR. pp. 398, August 1991. [4] Smart home research, Li Jiang, Da-YouLiu, Bo Yang, College of Computer Science andTechnology, Jilin University,pp 659, August 2004. [5] Smart home research, Li Jiang, Da-YouLiu, Bo Yang, College of Computer Science andTechnology, Jilin University,pp 660, August 2004.