Crowdsourcing solutions to 2D irregular strip packing problems from Internet workers
Crowdsourcing solutions to 2D irregular strip packing problems from Internet workers
复制标题
DOI：
10.1080/00207543.2015.1102355
复制
发表时间：
2016-07
影响因子：
9.2
通讯作者：
G. V. Annamalai Vasantha;A. Jagadeesan;J. Corney;A. Lynn;A. Agrawal
中科院分区：
文献类型：
--
作者：
G. V. Annamalai Vasantha;A. Jagadeesan;J. Corney;A. Lynn;A. Agrawal
许多工业过程需要在从原板材切割或冲压部件之前嵌套2D轮廓。尽管经过数十年的持续学术努力，算法解决方案仍然是次优的，并且产生的结果通常可以通过手动检查来改进。然而，互联网提供了前景的新的“人在循环”的方法来嵌套的问题，使用在线工人生产包装效率超出了目前的CAM软件包。为了研究这种方法的可行性，本文报道了在线工作人员参与六个标准基准数据集的交互嵌套的速度和效率。为了确保研究结果准确反映在线提供的许多不同劳动力的不同教育和社会背景，研究对象来自印度IT服务（即农村BPO）中心和苏格兰北方的家庭佣工网络。人类工作者的结果（即时间和包装效率）与商业CAM包的基线性能和最近的研究结果进行了对比。该论文的结论是，在线工人可以持续实现包装效率约4%，高于该项目建立的商业基线。除了描述在线工人嵌套组件的能力之外，结果还通过报告基准问题的新解决方案和演示评估最佳工人采用的包装策略的方法，为算法解决方案的开发做出了贡献。
Many industrial processes require the nesting of 2D profiles prior to the cutting, or stamping, of components from raw sheet material. Despite decades of sustained academic effort, algorithmic solutions are still sub-optimal and produce results that can frequently be improved by manual inspection. However, the Internet offers the prospect of novel ‘human-in-the-loop’ approaches to nesting problems that uses online workers to produce packing efficiencies beyond the reach of current CAM packages. To investigate the feasibility of such an approach, this paper reports on the speed and efficiency of online workers engaged in the interactive nesting of six standard benchmark data-sets. To ensure the results accurately characterise the diverse educational and social backgrounds of the many different labour forces available online, the study has been conducted with subjects based in both Indian IT service (i.e. Rural BPOs) centres and a network of homeworkers in Northern Scotland. The results (i.e. time and packing efficiency) of the human workers are contrasted with both the baseline performance of a commercial CAM package and recent research results. The paper concludes that online workers could consistently achieve packing efficiencies roughly 4% higher than the commercial based-line established by the project. Beyond characterising the abilities of online workers to nest components, the results also make a contribution to the development of algorithmic solutions by reporting new solutions to the benchmark problems and demonstrating methods for assessing the packing strategy employed by the best workers.