Making the smart built environment even smarter

Data has grown in volumes and varieties, computational processing more powerful, and cloud data storage more accessible and affordable. These developments have made it possible to produce models that can analyze bigger, more complex data and deliver faster, more accurate results.[1]

This may explain why Artificial intelligence (AI) is gaining popularity in recent years. Many consumers will have their first interaction with AI in the built environment through voice commands given to a smart home personal assistant device such as those by Amazon, Google or Apple.

What can we expect of AI in the built environment, going forward?

Leveraging AI

Smart cities and smart buildings around the world generate an incredible amount of data on a daily basis. Yet the data has not been systematically collected, stored, analyzed or leveraged to drive efficiencies or to meet sustainability goals.

While AI represents the broader concept of machines being able to carry out tasks in an intelligent way, machine learning is a current application of AI based on the idea that we can give machines access to data, and they can use that data to learn for themselves.

Machine learning (ML) is an application of artificial intelligence (AI) that allows systems to automatically learn and improve from exposure to more data without being explicitly programmed. In other words, ML focuses on the development of computer programs that can access data and use it to learn for themselves.[2]

As ML evolves, a class of semi-supervised learning has also been utilized. Such semi-supervised learning typically uses a large amount of input data but only a small amount of corresponding output data.

With the right algorithm, critical information – such as where the problems are, what’s causing them – can be culled from the data flood, and delivered when and where such insights are needed. With such advanced tools, organizations will be better equipped to identify opportunities and to resolve problems. 

AI and ML: A trust issue?

While AI and ML hold enormous promise, they are still maturing as technologies. Understandably, there are anxieties and concerns about them across economies and societies.

Data quality and quantity are vital considerations. Since ML requires a significant amount of data to train the algorithm, it follows that the quality of data input into the model is important, as higher quality data should enable better predictive capabilities. Obtaining sufficient quality labeled data can be a costly proposition and often depends on human experts to perform the labeling.

Trusting that the machine and data will make the right decision. As AI and ML become more advanced, they will start to make more sophisticated decisions. Some will question if automated processes could “learn” patterns that lead to undesired or unintended consequences or biases. If the underlying data set included biases, or is collected from a process that structurally included some form of bias, the ML algorithms will replicate and perpetuate those biases.

Hence, business leaders will need to pay careful attention to the history and provenance of their datasets. Many ML algorithms and resulting models are created as combinations of thousands of variables and their predictions are not easily explainable, which can make it difficult for observers to trust the output of the algorithm.

Will AI/ML take our jobs away? With AI/ML technologies becoming more ubiquitous, economic effects and impacts to the world of work are inevitable. As automation replaces many administrative, predictable physical tasks, data collection and processing, job functions will change. Concerns exist that many jobs as we know them today will be lost altogether or significantly re-shaped, at a minimum. Business leaders and policy makers need to take these concerns seriously and assume some responsibility for enabling and applying “Responsible AI.”

Machine learning in action

The next-generation smart buildings are about to become self-conscious, self-healing and occupant-driven buildings. Imagine empowering building occupants – such as, employees, visitors, doctors and patients – to interact with their environment for better comfort and productivity.

In the built environment, ML is foundational to many systems that rely on biometric recognition to enable physical access. For example, in addition to a physical access control credentials (such as, card or pin), cameras are used to verify users’ identity. The technology powering these cameras uses supervised machine learning techniques like neural networks to identify users. Modern AI algorithms can identify users with very high confidence, and the combination of facial recognition and traditional card access provides a higher level of assurance for secure access while minimizing disruption for the users.

ML is also being applied in energy management and predictive energy optimization in commercial settings. Corporate real estate owners and operators can leverage internal and external data to benchmark building performance, monitor building equipment, ensure occupant comfort and forecast operational budgets.

With predictive analytics, the technology can analyze load management (for heating, ventilation air-conditioning (HVAC), lighting, appliances and devices, in addition to fault detection and diagnosis. For instance, AI-based capability can predict anomalies in building equipment such as chillers, boilers, cooling tower or lighting, which allows potential issues to be addressed before something serious happens.

In addition, predictive analytics can also manage upcoming load, either by preemptively shedding load or optimizing processes to prepare for upcoming challenges. For example, intelligently pre-cool or store chilled water in advance of anticipated loads, creating just the right amount of chilled water early in the day before peak utility rates apply. The technology can also compute operational set-points for various components of HVAC systems with feedback of human comfort and energy consumption. With AI, the energy consumption of a whole building or even a piece of equipment can be predicted, which enables peak shaving, avoiding utility penalty, cost-effective energy supply planning and energy demand planning.

Voice control of building functions could be next on the table for commercial buildings. Currently, voice control is not yet big in commercial settings outside of personal devices like the building occupant’s phone, laptop or wearables. However, more organizations are beginning to deploy voice controls in communal spaces, such as in conference rooms. The take-up will largely depend on trust – such as the ease-of-use, how well the technology performs its intended use, and the protection of users’ privacy. Data has become a remarkable resource with truly amazing potential – for those who are able to gather it, understand its meanings, and put it to work. A smart strategy towards facilities management is a data-driven approach where building systems and equipment are linked to provide efficient, centralized control, and powered by AI/ML to mine the data – generated by the linked systems and supplied by external sources – for opportunities for improved efficiency and performance.