A Survey of Elevator Control Algorithms: From Classical Dispatching to Deep Reinforcement Learning with an Analytical Queueing Comparison

Nasir Ansari; Ritu Khanna; Saurabh Mehta

doi:https://doi.org/10.14445/22315373/IJMTT-V72I6P105

International Journal of Mathematics Trends and Technology

Research Article | Open Access | Download PDF

Volume 72 | Issue 6 | Year 2026 | Article Id. IJMTT-V72I6P105 | DOI : https://doi.org/10.14445/22315373/IJMTT-V72I6P105

A Survey of Elevator Control Algorithms: From Classical Dispatching to Deep Reinforcement Learning with an Analytical Queueing Comparison

Nasir Ansari, Ritu Khanna, Saurabh Mehta

Received	Revised	Accepted	Published
25 Apr 2026	30 May 2026	18 Jun 2026	30 Jun 2026

Citation :

Nasir Ansari, Ritu Khanna, Saurabh Mehta, "A Survey of Elevator Control Algorithms: From Classical Dispatching to Deep Reinforcement Learning with an Analytical Queueing Comparison," International Journal of Mathematics Trends and Technology (IJMTT), vol. 72, no. 6, pp. 45-58, 2026. Crossref, https://doi.org/10.14445/22315373/IJMTT-V72I6P105

Abstract

Elevator cars assignment rules regulate passengers' waiting time, energy consumption, and satisfaction in high-rises. This paper reviews ten algorithm families applied to solving the elevator dispatching problem through analysis of the development of solutions, starting with simple mechanical algorithms in the 1950s up until deep reinforcement learning in current days. Families of algorithms are classified into Classical and Deterministic Control, Computational Intelligence, and Learning-based and Hybrid Control categories. In particular, Classical and deterministic control covers such families as Single Automatic Operation, Selective Collective Operation, Group Control, and Destination Control. Computational Intelligence includes fuzzy logic, neural networks, genetic algorithms, and model predictive control. Lastly, Reinforcement learning and hybrid meta-control belong to the third group. For each family, unified notations of mathematics are used to describe the queueing models and theoretical framework. Literature choice follows PRISMA 2020 guidelines, and the area is defined with taxonomy, chronology, and comparison tables. To make a quantitative comparison of different approaches possible, a set of queueing models (M/M/1, M/M/c, and M/G/c) was studied using a realistic twenty-four hours calls distribution for a fifteen-storey building with five elevators, including uncertainty bands calculated through resampling. It should be emphasized that a comparison was made basing on queueing models, but not the discrete-event elevator trajectories simulation. Performance improvements associated with advanced algorithm families were estimated according to cited literature, and therefore, the results show relative trends, but not the measurements. The hierarchy was discovered, where the lowest waiting times belong to hybrid and learning-based algorithm families, while Single Automatic Operation exceeds the upper ISO 4190 limit during the period of the rush hour. Unresolved issues and future directions for research are outlined below.

Keywords

Elevator control, Group control, Destination dispatching, Queueing models, Reinforcement learning, Model predictive control, survey, analytical queueing comparison.

References

[1] M. Baraa, “A Formula for the Numerical Range of Elementary Operators,” ISRN Mathematics Analysis, vol. 2014, pp. 335-401, 2014.
[CrossRef] [Google Scholar] [Publisher Link]

[2] Benjamin Kimeu Daniel, Sammy Wabomba Musundi, and Kinyanjui Jeremiah Ndungu, “An Application of Maximal Numerical Range on Norm of Basic Elementary Operator in a Tensor Product,” Journal of Progressive Research in Mathematics, vol. 19, no. 1, pp. 73-81, 2022.
[Google Scholar] [Publisher Link]

[3] Denis Njue King’ang’I, “On Norm of Elementary Operator of Length Two,” International Journal of Scientific and Innovative Mathematical Research, vol. 5, no. 9, pp. 34-38, 2017.
[CrossRef] [Google Scholar] [Publisher Link]

[4] Martin Mathieu, “Elementary Operator on Calkin Algebras,” Irish Mathematical Society, vol. 46, pp. 33-42, 2001.
[Google Scholar] [Publisher Link]

[5] Peter G. Muiruri, Denis N. King'ang'i, and Sammy W. Musundi, “On the Norm of Basic Elementary Operator in a Tensor Product,” International Journal of Scientific and Innovation Mathematics Research, vol. 6, no. 6, pp. 15-22, 2018.
[Google Scholar] [Publisher Link]

[6] Peter Guchu Muiruri, “On the Norm of Elementary Operator of Length Two in Tensor Product of C*-Algebras,” International Journal of Mathematics, vol. 7, no. 8, pp. 1-9, 2024.
[CrossRef] [Google Scholar] [Publisher Link]

[7] Mohamed Barraa, and Mohamed Boumazgour, “A Lower Bound of the Norm of the Operator 𝑥 ⟼ 𝑎𝑥𝑏 + 𝑏𝑥𝑎,” Extracta Mathematicae, vol. 16, no. 2, pp. 223-227, 2001.
[Google Scholar] [Publisher Link]