Multiagent Approach for Real-Time Collision Avoidance and Path Replanning for Cranes

Abstract

Collisions on construction sites are one of the primary causes of fatal accidents. This paper proposes a multiagent-based approach to provide real-time support to the staff of construction projects. Collision avoidance is achieved by informing the crane operators about potential collisions and by providing motion replanning for crane operations. During the planning stage, a three-dimensional (3D) model of the static environment is created, and collision-free motion plans are generated by the agents for the cranes, considering engineering constraints and operation rules. During actual construction work, all mobile objects are tagged when entering the monitored area. A site state agent uses a real-time location system (RTLS), such as an ultra-wideband (UWB) system to collect location data, calculates the poses of the objects on site, and sends this information to other agents. By using this real-time updated information, agents can detect potential collisions and replan the path for the cranes for collision avoidance. A coordinator agent coordinates the movement of cranes by deciding their priorities. The site state agent, coordinator agent, and crane agents can communicate and negotiate with one another to make better decisions. The framework of the multiagent system is described in detail, and a prototype system is developed. Three case studies are used to verify and validate the proposed approach. The benefit of using the agent system is that real-time collision avoidance can be achieved by providing more awareness of the site situation and decision making through communication and negotiation between multiple agents, which results in safer and more productive work environment.