AI-Powered Elevator Design: Optimizing for People and Buildings

An elevator is one of those things you only notice when something goes wrong: you wait too long for it, it's overcrowded during peak hours, or, conversely, it sits idle in a building where almost no one lives above the third floor. It might seem like a minor issue. But from an engineer's perspective, behind every such inconvenience lies a decision made during the design phase – and one that usually fails to consider who will actually be using the elevator.

Researchers from the Moscow Institute of Physics and Technology (MIPT) have proposed a different approach. They have developed a system that uses artificial intelligence to design elevator equipment for a specific building and its residents, rather than relying on one-size-fits-all standards.

The Problem with Standard Elevator Design

The Root of the Problem

Traditionally, elevators are designed based on building codes and standard calculations: the number of floors, the number of apartments, and the estimated load. This works like a template – the same approach is applied to very different buildings.

The problem is that the real pattern of elevator usage almost never matches the calculated one. An elevator in a residential building with elderly residents is used differently than one in a student dormitory. In a business center, peak load occurs in the morning and evening, while in a hospital, it's spread almost evenly throughout the day. The template-based approach ignores all of this.

This is precisely where there is room for a smarter solution.

MIPT's AI Solution for Elevator Design

MIPT's Solution

The developed system analyzes the behavior of a building's occupants – how they use the elevator at different times of day, depending on the building type, number of floors, and traffic flow patterns. Based on this data, the AI helps determine the optimal parameters for the elevator system: the number of cars, their load capacity, speed, and dispatching algorithms.

Simply put, the system doesn't just count “how many people are in the building,” but models how those people move. Based on this model, it proposes a configuration that will operate efficiently under a building's specific conditions.

This is especially important during the design phase of a new building, when there is still an opportunity to set the right parameters – before the elevator shaft is constructed and the equipment is purchased.

Beyond Convenience: The Impact of Elevator Design

Why It's More Than Just Convenience

An elevator is not just about comfort. In high-rise buildings, it's a piece of infrastructure that determines housing accessibility for people with disabilities, elderly residents, and families with young children. A poorly designed system creates real problems for decades, and retrofitting it later is extremely expensive.

On the other hand, an oversized system is also a problem: extra cars, excessive load capacity, and more powerful motors. All of this increases the cost of both constructing and operating the building.

The AI system's task is to find a balance between “too little” and “too much,” relying on real-world data rather than standardized norms.

Applications of AI in Elevator Design

Potential Applications

The technology is primarily aimed at designers and developers – those who make decisions about elevator equipment before construction begins. However, the system could also be useful for retrofitting existing buildings when the elevator system needs an upgrade and the optimal configuration must be determined.

Interestingly, this approach – using AI not for content creation or communication, but for optimizing engineering solutions – is becoming increasingly common. Such systems are already being used in logistics, energy management, and urban infrastructure planning. Elevator design is yet another field where accumulated data on human behavior can be translated into specific engineering solutions.

Future Questions for AI-Powered Elevator Systems

Unanswered Questions

It is not yet entirely clear what data the system was trained on or how well it adapts to atypical scenarios – for example, mixed-use buildings with both residential and commercial floors. Another open question is how the system will behave as the resident demographics change over time: a building ages along with its inhabitants, and the load on the elevator system evolves.

This is not a criticism, but rather the natural questions that arise with any new engineering development. The answers will likely emerge as the system is put into practice.

The idea itself – designing infrastructure for people, rather than fitting people to the infrastructure – sounds obvious. But it is technologies like this one that help make it a reality, not just an idea on paper.