Orona Elevator Tech Leverages Quantum Extreme Learning Machine for Efficient Waiting Time Predictions

Orona, a global leader in elevator technology, has implemented Quantum Extreme Learning Machine (QELM) in its scheduling software to predict waiting times. The approach, known as QUELL, has shown significantly better prediction quality than traditional Machine Learning models. QUELL’s effectiveness does not diminish when using fewer features, making it a promising tool for various industries. Despite its potential, the application of QELM in real-world scenarios remains limited. This article explores the use of QELM in Orona’s elevator technology and discusses its potential for other industrial applications.

What is Quantum Extreme Learning Machine (QELM) and How is it Applied in the Industrial Context?

Quantum Extreme Learning Machine (QELM) is an emerging technique that uses quantum dynamics and an easy-training strategy to solve problems such as classification and regression efficiently. Despite its potential benefits, its real-world applications remain limited. This article presents an industrial application of QELM in the context of elevators, specifically in predicting the waiting time related to the scheduling software of elevators.

The scheduling software is implemented by Orona, a company globally recognized for its elevator technology. The approach, called QUELL, uses QELM to predict waiting times, with applications for software regression testing, elevator digital twins, and real-time performance prediction. The results show that QUELL can efficiently predict waiting times with prediction quality significantly better than that of classical Machine Learning (ML) models employed in a state-of-the-practice approach.

Moreover, the prediction quality of QUELL does not degrade when using fewer features. This article also provides insights into using QELM in other applications in Orona and discusses how QELM could be applied to other industrial applications.

How Does Orona Utilize QELM in Elevator Technology?

Orona, a Spanish company known for building elevators, uses a scheduling software called the dispatcher at the core of its elevators. This software assigns an elevator to each call while maintaining an acceptable quality of service (QoS). A commonly used metric for measuring the QoS of the elevators is the Average Waiting Time (AWT), which is the average time passengers have to wait from when they press the call button until the time the elevator arrives.

The dispatcher software undergoes evolution like any other software system, requiring regression testing. In Orona, a regression test oracle is employed during design time in both Software in the Loop (SiL) and Hardware in the Loop (HiL) settings. To reduce testing costs, this regression test oracle was proposed to be replaced by a machine learning (ML)-based test oracle.

As the DevOps paradigm emerges in the context of Cyber-Physical Systems (CPSs), Orona is interested in deploying such oracles also in operation to see whether the QoS of the elevators meets acceptable values. However, the application of existing ML-based oracles is not straightforward.

What are the Challenges and Solutions in Applying ML-based Oracles in Orona?

During design time, several features are available. For example, the passengers’ weight is a feature that an ML-based test oracle can use at design time to determine whether the AWT of passengers is within an acceptable threshold in a given time window. However, some of these features, such as the passengers’ weight itself, are not easy to measure at operation time, hence they cannot be used as features for ML models to be used at operation time.

Moreover, elevators must be configured according to the constraints imposed by the building where they are installed. As a result, a different number of features are available for different elevator installations. Hence, in Orona, there is an increasing need for ML-based test oracles that can be trained on a variable number of features, depending on the installation configurations.

To cater to these needs, Orona explores the use of Quantum Extreme Learning Machine (QELM) by proposing a QELM-based approach called QUantum Extreme Learning eLevator (QUELL). QELM is a quantum machine learning technique that uses quantum dynamics of quantum reservoirs to enable a simple machine learning model to be efficiently trained with a limited number of features but still provide good prediction quality.

How Effective is QUELL in Predicting AWT?

To assess the effectiveness and efficiency of QUELL, operational data from four days of operation of a real elevator installation was used. The results showed that QUELL can significantly outperform the classical ML-based regression models for the prediction task, thereby demonstrating QELM’s potential benefits in an industrial context.

Based on these results, the application of QUELL to various contexts at Orona is discussed, followed by a discussion on the potential applications in other industrial contexts valuable for practitioners. Finally, open research questions related to QELM applications in software engineering that are valuable for software engineering researchers are presented.

What is the Future of QELM in Industrial Applications?

The potential of QELM in industrial applications is vast. Its ability to efficiently train with a limited number of features while still providing good prediction quality makes it a promising tool for various industries. In the context of Orona, the application of QUELL has shown significant improvements in predicting the Average Waiting Time (AWT) of elevators.

This success opens up possibilities for its application in other areas within Orona and in other industrial contexts. However, more research is needed to fully understand and maximize the potential of QELM in industrial applications. The open research questions presented in this article provide a starting point for further exploration into the applications of QELM in software engineering.