Smart Living: The Role of AI in Making Homes Smarter (redefines the concept of home automation)

Preface

Motion sensing light switches have been around for over 30 years.

Code mandates enforced early deployments of these switches in commercial environments like restrooms. Though the primary use case was energy savings, it eventually became a generalized convenience. Someone entering a restroom didn’t need to hunt for the switch in the dark. On leaving they didn’t need to turn off the light or concern themselves if there was someone still in the restroom.

As code mandates were amended for energy savings, vacancy switches became mandatory for certain residential areas. As people started experiencing motion sensor switches, they have been introduced to this initial and rudimentary level of automation.

Limitation of Motion and IoT Switches

Vacancy style motion switches were never intended for general home automation and cannot be deployed in most living spaces because of the lack intelligence and the limited and physical nature of changing settings (such as sensitivity and duration).

With the advent of IoT devices, regular light switches have become a new consumer toy by providing remote control via app and/or voice. The consumer still needs to command the switch to operate, so this cannot be deemed as true automation.

Unfortunately, vacancy switches that meet code requirements generally lack IoT technology. While many IoT switches have not embraced motion or other sensing and relied on just providing connectivity and requiring app/voice control.

An advanced light switch that can provide vacancy compliance and also to have the intelligence to provide automation in various rooms and areas of the home requires advanced software.

The experience we want

We would like our home to understand our needs. Like, walking into a kitchen with grocery bags in both hands in the evening and the lights coming on automatically; like those mandated by code in restrooms.

But the lights would not need to come on during the daytime when there is sufficient natural light or come on at a dim level late at night when doing a fridge snack run.

This is automation that is aware of the different times of the day, the present room conditions and the expectations of the occupant.

The Solution

To achieve this more advanced automation, sensors are required, though more importantly, intelligence at the edge (in the switch itself) is needed.

One area that AI has recently come to the forefront is in chatGPT, a language model AI that correlates crawled / indexed data with an inquiry that then allows the AI engine to start generating an answer. The answer is similar to how a human mind would start processing a question, comprehend and start answering it.

Sensing
Similar to humans having 5 sensory inputs, for a holistic home automation solution, multiple and continuous sensing data availability throughout the home is necessary.
And this is where the struggle begins. Name one product that can provide such a continuous data stream throughout the home.
Even if such a product was available, where is this data going be stored, analyzed, and processed? There is a need for intelligence to make decisions to perform the operation the person would typically do manually. This new touchless experience is true automation and the goal.
When the BriteHome founding team was faced with finding a solution, we came up with a strategy of building a reference hardware design utilizing a multi-sensor framework. Sensors include activity / occupancy / presence, ambient light, power metering, temperature, humidity, indoor air quality, barometric pressure, and CO2 equivalent to help obtain the relevant home automation data.

Positioning
The next task was determining the ideal product that would utilize this reference sensor hardware to capture data throughout the home without invading the occupants’ privacy.
We studied various products that are common to every room, like light bulbs, lamps, pucks, and light switches. We selected the light switch; it is always powered and is a ubiquitous product in every room of the home. A product that is familiar to all generations.
Having chosen the light switch as the product, sensor and processing requirements identified, the engineering team could begin the challenging task of fitting everything into a standard form factor product. A major design criterion was ensuring that the final cost of the light switch was well positioned for market acceptance.

Software

With the hardware base ready, the software team focused on its core competency, to architect and implement the software intelligence functionality. Several software components were needed to create a finished product, and these included the firmware running on the switch itself, high level analytics and processing on the cloud and finally the smart phone App.

AI – The automation Intelligence Software

Leveraging the granular data generated by the BriteHome solution, supervised learning helped us mine patterns and derive AI models specifically for home automation usage. With rich data now available the possibilities of providing true AI automation is immense.
Examples:
- Pattern mining: Occupant’s preference on light settings for each living space in home throughout the day.
  Application: Automated lighting control configuration
  The first field application of AI for lighting control is already available in the BriteHome App, where a BriteLight switch auto configures the daily routines based on its location within the home. This model was developed via supervised learned patterns of occupants’ preferences.
  The advanced model can have distributed intelligence from the edge through to the cloud so that thereare no critical paths in day-to-day functionality for lighting control.
- Pattern mining: Occupancy for every living space within the home throughout the day.
  Application: DIY contextual security
  Room-by-room holistic view of the home occupancy. When the home is fully vacant, the system can auto arm without need to punch in codes or confirm with an app. The home can disarm as the homeowner arrives home.
  At night the system can auto arm the downstairs floor, if all occupants are sleeping upstairs.
  Both of above models were developed via supervised learned patterns of occupancy in each room of the home.
  Utilizing historical data, these models can enforce polices for both security as well as vacancy mode. In vacancy mode, lights can operate and provide a realistic lived-in appearance when the home is vacant.
  The next level of the AI model will be unsupervised and more advanced using logistics and linear regression, and neural networks to predict when the home will be vacant or occupants going on a vacation (longer vacancy of the home).
  Application: Aging in place
  It is even more intriguing when occupancy data is fed to supervised AI models to correlate movement patterns between areas of the home to learn typical daily routines. These patterns enable the differentiation between normal and abnormal behavior. e.g., waking up and going from the bedroom to the bathroom, but then detect that the routine pattern of occupancy in the kitchen is missing. Even long-term activity patterns can be monitored to spot gradual degradation of activity levels which can be correlated with certain kinds of health degradations.

Summarizing

In the realm of home automation, meaningful data and applied intelligence are not distant concepts. As we begin to witness how AI can resolve the challenges of complex automation and services, the apprehension about navigating complex home automation systems and related services is likely to dissipate as users begin to experience its benefits.