Construction faces significant challenges, including high incident rates and shortage of skilled labor. With the sector’s increasing willingness to integrate robotics to improve productivity, this paper addresses the emerging safety challenges in human–robot collaboration (HRC). We present a novel approach that leverages human feedback regarding robot behaviors to define safety-oriented control actions. By training a preference prediction model with human input, we demonstrate the effectiveness of our approach in guiding robots towards safer behaviors in complex construction tasks. The approach begins with designing an online labeling tool tailored for collecting human preference data regarding robot behaviors in collaborative tasks. A score model is then trained to enable prioritization of safer robot behaviors. Finally, safety-oriented robot behaviors can be inferred. This research underscores the importance of aligning construction robot behaviors with human preferences, offering a scalable solution to enhance occupational safety for construction.

The construction industry necessitates complex interactions among multiple agents (e.g., workers and robots) for efficient task execution. In this paper, we present a framework that aims at achieving multiagent reinforcement learning (RL) for robot control in construction tasks. Our proposed framework leverages the principles of proximal policy optimization (PPO) and develops a multiagent variant to enable robots to acquire sophisticated control policies. We evaluated the effectiveness of our framework through four collaborative construction tasks. The results revealed the efficient collaboration mechanism between agents and demonstrated the ability of our approach to enable multiple robots to learn and adapt their behaviors in complex and dynamic construction tasks while effectively preventing collisions. The results also revealed the advantages of combining RL and inverse kinematics (IK) in enabling precise installation. The findings from this research contribute to the advancement of multiagent RL in the domain of construction robotics. By enabling robots to collaborate effectively, we pave the way for more efficient, flexible, and intelligent construction processes.