Construction projects are complicated, and their success relies heavily on various factors
[1]. Traditionally, construction projects have faced numerous challenges, including delays, cost overruns, and safety concerns. These issues often arise due to human error, inefficient resource allocation, and inadequate planning. Similarly, human activities become efficient in smart and sustainable communities
[2]. Artificial Intelligence (AI) and Machine Learning (ML) have the potential to significantly enhance construction processes and contribute to the development of sustainable communities. Specifically, AI has the potential to increase labour efficiency by 40% and quadruple yearly economic growth rates by 2035
[3]. AI, a branch of computer science, focuses on creating intelligent machines capable of performing tasks that ordinarily need human intelligence. AI can find non-obvious patterns in data while also producing reliable forecasts of the expected future in previously unexplored circumstances
[4]. ML is a key branch of AI
[2] that allows a computer to learn from data, uncover patterns, and ultimately make judgments and predictions with minimal human intervention
[5].
Emerging AI techniques, such as artificial neural networks (ANNs), can be used to generalize hidden patterns and implicit associations from historical data, resulting in a viable prediction model to assist the planner in analysing new cases in the issue area
[4]. Artificial neural networks are a class of ML algorithms capable of modelling nonlinear relationships between input vectors and target values
[7]. More so, digital twins (DTs) integrate AI, ML, and data analytics to create living digital simulation models that can learn and update from multiple sources as well as represent and predict the current and future conditions of physical counterparts
[8]. Residential and commercial buildings account for a significant portion of global energy consumption; therefore, hourly predictions of electricity consumption in residential and commercial buildings are required to support operational decisions, demand response strategies, and the installation of distributed generation systems
[7]. Thermal comfort is a key component of smart building control and operation, as well as building design and modelling
[9]. Also, the building and construction sectors consume one-third of the total world’s final energy consumption and emit roughly 15% of CO
2 [10].
2. Indoor Environment
2.1. Energy Management
Ref.
[11] employed AI to turn the Europoint Towers in Rotterdam into self-sufficient buildings by taking into account energy use and food production (lettuce crops). The study looked into optimizing high-rise buildings for self-sufficiency in food production and energy usage based on daylight availability. Unlike the majority of AI models now used in energy forecasting, which is traditional and deterministic, ‘transformer,’ a novel deep learning paradigm, leverages the idea of self-attention
[12]. The study developed a transformer-based model to predict the energy consumption of a real-world university library and compare it to a baseline model. Ref.
[13] delved into pilot systems and prototypes that demonstrate how AI may aid in the process of achieving energy sustainability in smart cities. The study investigated smart metering and non-intrusive load monitoring (NILM) to establish a case for the latter’s utility in profiling electric appliance power usage. Using ML approaches, Ref.
[14] investigated the energy consumption trends of residential assessment units. Ref.
[15] focused on creating an energy management approach that combined photovoltaics and storage systems, using a multi-story building with a high density of families as the major case study to provide data that allows feasibility forecasting. Ref.
[16] aimed to reduce the computation required to determine the energy consumption of various combinations by identifying suitable training samples, computing their energy consumption with EnergyPlus, and estimating the rest of the data’s energy consumption with ML techniques.
Based on the experts’ competence, ref.
[17] attempted to analyse the electrical energy usage in Mashhad, Iran, using ML methods to offer dynamic solutions for encouraging residents’ interest in renewable energy generation. Ref.
[18] used machine learning interpretability approaches to predict whether a room is occupied or unoccupied, resulting in energy savings in buildings. Ref.
[19] developed an AI-based framework for addressing various scientific issues in green buildings, such as providing clean energy, developing a smart and sustainable biogas production control system, and integrating solid waste management with the Sustainable Development Goals (SDGs). The AI techniques used were Random Tree (RT), Random Forest (RF), artificial neural network, and Adaptive-Network-based Fuzzy Inference System (ANFIS). Ref.
[20] provided a bottom-up strategy for creating heat load analysis and forecasting utilizing ML techniques such as support vector machines, feed-forward neural networks, multiple linear regression, and regression trees.
[21] unveiled a new simulation environment created by combining CitySim, a building energy simulator, and TensorFlow, a powerful ML library capable of developing building energy scenarios in which ML algorithms are applied to the major problems and opportunities that modern cities face. Ref.
[22] developed a novel ensemble model based on actual data to estimate energy consumption in residential buildings.
The ensemble model combines two supervised learning machines—least squares support vector regression and the radial basis function neural network—and incorporates symbiotic organism search to automatically discover its best tuning parameters. The study by
[23] developed a technique for estimating domestic energy demand that included statistical data matching, ML, and household/population synthesis. Through the combination of data-driven methodologies with physics-based model ML algorithms, ref.
[24] created a hybrid model to handle the problem of residential electricity consumption forecasting. Using an AI model, ref.
[10] forecasted residential building energy use and greenhouse gas emissions. Ref.
[25] addressed the need to develop methods for accurately modelling and characterizing building energy consumption in cities by proposing a novel data-driven urban energy Simulation (DUE-S) framework that integrates a network-based ML algorithm with engineering simulations to better understand how buildings consume energy across multiple temporal and spatial scales in a city.
2.2. Thermal Comfort, Power, and Cooling
Ref.
[26] built an automated platform that offers data on home power consumption in each Taiwanese city. To anticipate future domestic electricity consumption, ML was employed. Furthermore, to improve the accuracy of the top machine learner, a nature-inspired optimization strategy was used, resulting in an even superior hybrid ensemble model. The suggested approach in
[27] is a hybrid of artificial neural networks and stochastic fractal search (SFS-ANNs) designed to handle the problem of early cooling demand prediction in buildings. To forecast indoor temperature more correctly and effectively, Ref.
[28] presented a hybrid model based on feature selection approaches such as feature significance and support vector regression (SVR). Ref.
[7] created a recurrent neural network (RNN) model with a one-hour resolution to produce medium-to-long-term predictions of power consumption profiles in commercial and residential buildings. Ref
[9] employed ML to bridge the gap between controlled building factors and thermal comfort. The study demonstrated that neural networks are good ML approaches for simulating comfort levels.
2.3. Circulation and Automation
The study
[29] takes advantage of the image collecting and processing system’s knowledge of passenger group sizes and waiting durations. The objective was to create a decision engine that could govern the elevator’s movements while increasing user satisfaction. Ref.
[30] investigated edge AI-enabled technology and suggested a fully featured IoT and edge computing-based cohesive system for smart home automation.
3. Outdoor Environment
3.1. Pollution—Air, Noise, and Waste
Ref.
[31]’s study on AI-based air quality early warning systems is expected to play a vital role in its future accuracy and usefulness. Ref.
[32] suggested an improved ML strategy for predicting urban ambient particulate matter (PM
2.5) concentrations that combines cascade and PCA algorithms to reduce data dimensionality and investigate nonlinear relationships across variables. Ref.
[33] proposed a novel algorithm based on cloud model granulation (CMG) for air quality forecasting. Ref.
[34] developed a system that monitors and forecasts urban air pollution by using ML algorithms to construct credible forecasting models for various air pollutant concentrations. Ref.
[35] suggested a network that predicts future air quality, resulting in cutting-edge performance in urban air quality prediction. Ref.
[36] presented an ML technique based on six years of meteorological and pollutant data analysis to forecast PM
2.5 concentrations from wind (speed and direction) and precipitation levels. Ref.
[37] presented a cost-effective and efficient air quality modelling framework that incorporates various elements while utilizing cutting-edge AI-based approaches.
Using environmental monitoring data and meteorological observations, Ref.
[38] developed an ML-based strategy for reliably predicting the air quality index. Ref.
[39] created a regression model of daily air quality forecast using the SVM approach at the local scale in the Gijón metropolitan region of Northern Spain. Ref.
[40] investigated a new technique of daily air pollution prediction based on observed carbon mono oxide (CO) concentrations utilizing a combination of Support Vector Machine (SVM) as a predictor and Partial Least Square (PLS) as a data selection tool.
Ref.
[2] created a rule-based ML model to assess the influence of city and nation variables on the disposal of waste. The findings identified municipal government, employment, and technical research as key factors influencing sustainable waste management. To choose waste-to-energy plants, Ref.
[41] created and used a hybrid framework that included the analytical hierarchy method with ML approaches. Ref.
[42] offered an investigation of three AI-related models as tools for forecasting the development of urban solid waste in the city of Bogota to learn the behaviour of such types of waste.
Renaud
[43] investigated the capacity of Gradient Boosting and Deep Learning to produce long-term noise level forecasts using noise data gathered in a suburb of an English metropolis and then offered a strategy for identifying noise level anomalies based on predictions.
3.2. Real Estate and Prices
Ref.
[44] offered an overview of ML approaches for forecasting property values. Ref.
[45] provided an experiment on estimating real estate prices using seven ML approaches and five years of historical data on real estate transactions in major French cities. A unique ML approach was presented in
[46] to address the complexity of real estate modelling. The study investigated the possibility of call detail data for forecasting real estate prices using AI. Ref.
[47] used ML approaches to forecast house prices in two Italian cities. Ref.
[48] developed an innovative and complete model for calculating the price of new houses during the design or early construction phase by combining a deep belief-restricted Boltzmann machine with a unique non-mating genetic algorithm. Ref.
[49] employed location-based services APIs as an urban data source to assess the attractiveness of a residential area for users looking for long-term rental apartments by developing a machine learning model to forecast days on the market. As a research approach, ML algorithms were employed in the study
[50] to construct a house price forecast model.
3.3. Infrastructure Development
Sousa et al.
[51] combined semantic modeling and data-driven AI methodologies to deliver autonomous assessments for the operation of a public street lighting network to maximize energy usage while maintaining light quality patterns. To optimize the waste management process, ref.
[52] presented an AI-based Hybridized Intelligent Framework (AIHIF) for automated recycling. The system introduced using ML and graph theory will maximize waste collection within a limited distance. Ref.
[53] suggested an ML-based technique that could be utilized to extract elements of regional architectures and assess architectural forms in the process of urban redevelopment. Ref.
[54] offered a novel AI-based technique for an urban-scale application that quantifies both subjective and objective human-scale streetscape perceived quality. Ref.
[55] used ML algorithms to create models to aid in quick decision-making for optimal resource allocation in the aftermath of disruption and to assist investment decisions for the structural reconfiguration of urban systems.
Ref.
[56] created a unique hybrid AI model that predicted building destruction in South Korean redevelopment zones by combining standalone algorithms with architectural and engineering technologies. Ref.
[57] demonstrated how the multi-zone optimization (MUZO) methodology developed in the first phase of their research project could improve the overall performance of a high-rise building in crowded metropolitan neighbourhoods. Ref.
[58] explored the various challenges in the formulation and execution of overall country-specific urban planning by combining big data technology and ML to build a virtual design model of urban planning and develop the functional structure of the model based on actual demands. Ref.
[59] presented the MUZO methodology that supports decision-making for high-rise buildings per floor levels and performance aspects.
3.4. Life Cycle Assessment and Rainfall Prediction
Koyamparambath et al.
[60] explored AI techniques to forecast the environmental performance of a product or service per life cycle assessment (LCA). The data is processed using natural language processing (NLP), which is then taught to the random forest method, an ensemble tree-based machine learning approach. Ref.
[61] combined fuzzy cognitive maps with a metaheuristics-based rainfall prediction system (FCMM-RPS). The FCMM-RPS approach aims to predict rainfall in an automated and efficient manner.
3.5. Other Applications
Bui et al.
[62] offered an ML approach to replace traditional testing for determining the coefficient of soil compression. The novel method combines the Multi-Layer Perceptron Neural Network (MLP Neural Nets) with Particle Swarm Optimization (PSO). Using AI/ML approaches, ref.
[63] focused on environmental injustice. Ref.
[64] investigated a metamodel-based method that included simulated data gathering and data-driven approaches for forecasting and optimizing heating and cooling loads in three different climates in Morocco.