Seyedali Mirmotalebi - Civil, Architectural Engineering (CAE),
Ahad Alvandi - Electrical and Computer Engineering, Kharazmi University
Somia Molaei - Software Engineering, Iran University of Industries & Mines
Amirhossein Rouhbakhsh - Electronic Engineering, Northwestern Polytechnic University
Amirhossein Amadeh - Independent Researcher,
Ali Ahmadi - Civil Engineering, Iran University of Science and Tech

Various activities involving one or more body motions are part of construction tasks. Understanding the constantly shifting attitudes and behaviors of construction workers is crucial to effective management of construction workers is important for both their safety and productivity. Even though various research projects have produced encouraging activity recognition findings, more study is still required to determine the ideal places for motion sensors to be placed on a worker's body in order to increase performance and lower implementation costs. This study suggests evaluating several motion sensors mounted to workers doing common construction activities using simulation. The entire body's motion sensor data is gathered using a set of Inertial Measurement Unit (IMU) sensors. By simulating scenarios with various combinations and attributes of the sensors, multiple machine learning algorithms are used to categorize the motions of the workers. Each IMU sensor, which is installed in various body places, is put through simulations to assess its accuracy in identifying the various activity kinds of the worker. The performance of activity recognition is then examined in relation to the effectiveness of sensor sites to establish the relative benefit of each place. Based on the findings, the necessary number of sensors can be decreased while retaining the performance of recognition

IMU Sensors, Machine Learning, Deep Learning, Construction Workers, Motion