We're not in Kansas anymore
The traditional house with the white-picket fence is changing. The white picket fence is still here--but now it might be wired, and could let its owner know when you have opened the gate, walked up to the door, and wiped your feet on the mat. In fact, by the time you're lifting your hand to knock, the owner could have gotten an email or telephone call from his house telling him to expect company. The owner, who might be at home, on vacation or at work, could call the police or let you in-depending on the picture the video camera has taken of you and sent him. This isn't the future, it's happening right now.
Yi-Min Wang, a researcher in the Systems and Networking group at Microsoft Research, is the architect and owner of just such a system. His team, which includes Wilf Russell and Anish Arora, built the system and installed one in Wang's house and one in the lab. Wang was spurred to action by his pet fish. Wang lives in an area that experiences power outages, and he worried that his five fish tanks might suffer if the light, heat and filtration systems lost power. Wang started wondering how he could build a system that used off-the-shelf sensors and technology to protect his home environment.
"We didn't want to wait until the future smart appliances came out. So we took existing devices, very cheap consumer devices, and built the infrastructure to integrate them." Wang's system, called Aladdin, is controlled by six PCs operating on Windows 98. The system controls his garage door; sends him email or calls his cell phone when his power goes out or the water sensors in his crawl space go off; monitors his safe; lets him know when his wife has arrived home; sends him pictures of his front door and garage door; adjusts the house temperature; and turns the lights on and off in response to email messages he sends from any place that has an Internet connection.
As reliable as your telephone
Wang says his team wanted to build an ultra reliable system. "I don't want a system that is not reliable. If it only works 95% of the time, and I have to suffer the other 5% of the time, I don't want it." What if your telephone only worked 95% of the time? In a sense, home systems must be more reliable than business systems, Wang claims. A business will lose money if its system goes down, so it hires a systems administrator to watch over and fix any problems. However, the average consumer isn't about to hire an administrator to monitor his home system-he simply won't buy something that needs handholding from an expensive service person.
The team had to build a system that could handle any problem-even ones that they couldn't think of. Wang says he has discovered that you can't think of everything, so you must build a system that is smart enough to self-stabilize. Usually self-stabilization is talked about with a particular protocol, but this system needs self-stabilization at every layer, so that if anything happens it will get corrected and the system can continue to function.
Because this system functions over power and phone lines and through the Internet, problems with telephone systems, Internet access services and electrical power all have to be considered. If there's a power outage, the system must be set up to function for a long enough period of time after the outage to alert the owner to the problem, and power itself back up when the power returns. Wang has an uninterruptible power supply (UPS) that will operate his residential gateway computer for 30 minutes after an outage. He then has the choice to go home and save his fish, or wait for a message from his computer saying power has returned and all systems are functioning.
Home networking systems will fail more often because they contain devices from different manufacturers, are connected to different networks, and run different protocols. Wang's group considered this by using a soft-state store in the system. The hardware and software objects connected to the system periodically announce their existence and states by sending a signal or heartbeat to the soft-state store. Each object has a corresponding soft-state variable in the soft-state store, and every time the object sends a heartbeat the variable gets refreshed and the timer is reset. If the heartbeats stop because the announcing device, sensor or object fails, runs out of batteries or gets disconnected, the variable won't be refreshed and will time out. Changes in the soft-state store will be spread throughout the system, so that all the objects and applications can adapt.
As dependable as taxes
With all these emails flying back and forth, you might wonder if you're opening your home to cyber burglars. Wang thought of this too, and taught his system never to speak to strangers. All the emails are encrypted and digitally signed. There's also an embedded password for additional security. You can add another layer of security by customizing the language you use to give commands. For instance, instead of typing, "close the garage door," you could set up the system to only accept the command, "close the Van door."
For your added peace of mind, the system can send two emails to confirm its actions-since one email could get lost. If you've set your system to phone your cell or work phone when a sensor goes off, it will call you twice. If there's a critical command that you want to be extra sure has been carried out, you can have cameras installed to take a picture. Wang has cameras in his garage. When he sends a command to close the door, the system sends him before and after pictures to confirm the action. Wang says that consumer sensors sometimes have problems, and this is the only way to be sure. "I want to see it to believe it."
Wang also observed that, as the house gets smarter, the unusual glitches may get trickier. If the system detects that the device has gone bad and is sending random signals, or there's radio interference that gets translated into bogus signals, it needs to alert the owner to the problem. Wang's team has built a system that constantly monitors the power line for bad command patterns by matching them against mathematical expressions. Another way to address the problems of sensor instability is to install multiple sensors. For example, Wang installed three inexpensive sensors on his garage door. If any one of them has gone bad and is lying about the state of the garage door, the other two sensors can vote it out and notify the owner that there is a discrepancy.
The magic lamp
The Aladdin Device Adapter is the magic lamp in the system. The Adapter can take everyday, dull household appliances and make them smart. Wang demonstrated this in the lab system by plugging an ordinary lamp into the Adapter, which is plugged into an outlet that the system recognizes as the second floor master bedroom outlet. The Adapter announced to the system that the lamp is now available in the master bedroom. Wang then typed a natural language command into his laptop: "Turn on the light on the second floor master bedroom," and sent it by email to the lab system. The light turned on. The Adapter can also detect that the lamp has been physically switched off or the bulb has burnt out, and notify the system that the lamp is no longer available. The Adapter also sends heartbeats on behalf of the lamp, so that if both the lamp and the Adapter are yanked out of the outlet, the system will notice that the lamp is gone.
Consumers will come up with their own killer apps. Some will want to monitor the pet door to see what the cat has dragged in. Others might want to watch over their school-age children or an elderly parent. Some have basements or crawl spaces that have a tendency to flood; while others might want to find out which neighbor's dog has been digging up their carefully planted flowerbeds. Wang's system will allow consumers to configure their own home monitoring systems. They'll be able to pick up a sensor at the local store and plug it into the system, or easily add a device to the system with the Adapter. "This kind of extensibility, together with the reliability and dependability, are the keys to successful home networking." Wang concluded.