Worker-Aware Task Planning for Construction Robots: A Physiologically Based Communication Channel Interface

Unsafe workplace environment, labor shortage, and low productivity rate are among the most critical challenges faced by the construction industry. These ongoing challenges have prompted the construction research community to explore robotization as a new direction of improvement. Most of these endeavors have focused on human-robot collaboration (HRC) as the design and development of adaptable and scalable robots for the dynamic work environment of construction sites are excessively difficult. Although HRC can facilitate the design and implementation process, it can pose new threats to human workers’ physical and mental health. One promising approach for a safe and productive HRC is that the co-workers’ physiological responses are captured, interpreted, and infused as a new sensory information level into robotics optimization and planning systems. However, strict compliance of the robot to the system’s prediction results may lead to unstable robotic adjustments because the robot modifies its performance by blindly following the imperfect system information. The present research seeks to optimize such reliance through physiologically aware signal classification algorithms that can appropriately establish harmony between worker physiology and robotic performance. The backbone of this algorithm is grounded in the fact that human physiological alterations are relatively gradual as opposed to electronic alterations of the systems that can happen in a fracture of a second. The proposed algorithm innovatively screens the machine-generated prediction results so as to produce logically sound decisions for robotic control and manipulations. To validate the algorithm in a real robotic system, the authors examined its performance in a collaboration between an unmanned terrestrial robot and a construction worker, with the total of eight subjects in the study. The terrestrial robot was designed to capture physiological signals from the workers’ wearable biosensors, promptly translate them into high-level information, and modify its performance based on trained ML classifiers. The experiment results demonstrated the potentials of the proposed screening algorithm to improve the probability of the correct system decisions in adjusting the robot’s performance. This research shows the algorithm can facilitate the stable implementation of robots within physiologically based HRC at construction sites. This algorithmic filtering process can also be extended to human-system interactions that can facilitate the design and development of interfaces between workers and computer systems.