The development and evaluation of an on-site construction and demolition waste recycling robot

Abstract

Worldwide, the demolition and construction waste generated by construction activities has caused a large number of environmental pollution problems. Although more and more countries are currently adopting laws and regulations to encourage manufacturers to increase the recycling rate of demolition and construction waste, it has actually achieved little effect for various reasons. Many builders still send mixed construction and demolition waste to landfills for disposal. The purpose of this study is to develop a robot for sorting and picking up waste for construction and demolition, and to evaluate the feasibility of the robot prototype in a real environment. In order to achieve these goals, we have designed and evaluated the robot patrol system and the pick-up system separately, in order to achieve the following objectives: (1) to realize the automatic patrol of the robot prototype in complex construction sites; (2) to realize the automatic sorting and recycling of heterosexual construction and demolition waste Pick up; (3) confirm the feasibility of the robot prototype. First of all, due to the less application of robot path planning in the construction industry, we built a robot platform suitable for use on construction sites. Ensure the accuracy of the robot through technologies such as map segmentation and relocation. An advanced path planning algorithm is also used to ensure the efficiency of robot patrol tasks in any environment. Secondly, in order to automatically collect demolition and construction waste scattered on the ground, this study introduced a computer vision algorithm. However, there is currently no database for demolition and construction waste, so we built and expanded the database of target objects based on COCO format. Through experiment and optimization, the pixel-level target recognition system is completed. Thirdly, we first developed strategies for picking up demolition wastes of different shapes. In this study, we realized that the usual picking strategies sometimes produce errors when faced with small objects, and the success rate for special-shaped construction waste is even lower. Therefore, we have developed a set of picking strategies for elongated objects and curved objects, which can pick up water pipes and cables very well. The strategy can also be applied to other similar objects. Fourth, considering that the testing of robot prototypes in the laboratory does not well evaluate the feasibility of the real environment, an evaluation was conducted to investigate the accuracy and success rate of robot patrolling, positioning, target detection, and object pickup. The research shows that the prototype of the manufactured robot can be accurately located in the real environment, complete the patrol task and establish a real-time point cloud map, which is conducive to the management work on the construction site. In addition, in order to improve the accuracy of the computer vision system, different CNN backbones and transfer learning sources are compared and verified. Overall, this research work provides an innovative method of collecting construction and demolition waste to solve the current problems encountered in the recycling of construction and demolition waste and increase its recovery rate.