Occupancy sensors have long been used for control of various devices (like artificial light, HVAC devices, etc.). Past research has shown that use of occupancy sensors for control of lighting can save up to 30% electrical energy used for lighting. However, most of the sensors, which are currently being put to use, have a preset (sometimes user adjustable) time delay (TD) (the time after which the lights or any other load will be switched "off" after the last motion is detected by the sensor). If this TD is long then, it will have less energy savings, as the load will remain "on" during unoccupied period also. At the same time, if the To is kept short then it may result in unwanted switching "off" (False Off) of the lights when no motion is detected during periods of occupancy. It has been observed in our research that the activity level of a user changes over the time of the day. Also, it is seen that activity level of different users, is different. Hence, single TD for all the users and for all times of the day is not desirable. The commonly used sensors do not adapt to changing activity levels. In this paper, we present design of smart occupancy sensors which can adapt to changing activity levels. A model is also proposed for ＇＇human movement＇＇ of a person working at a computer. Smart occupancy sensor can learn the variation in activity level of the occupants with respect to time of the day. With this information, it can change the TD with time of the day. Experiments conducted have shown that about 5% more energy can be saved by using smart occupancy sensor as compared to non-adapting fixed TD sensors. (C) 2000 Elsevier Science S.A. All rights reserved.