Using RGB-D sensors and evolutionary algorithms for the optimization of workstation layouts Jose Antonio Diego-Mas a, * , Rocio Poveda-Bautista b , Diana Garzon-Leal c a I3B, Institute for Research and Innovation in Bioengineering, Universitat Polit ecnica de Val encia, Camino de Vera s/n, 46022 Valencia, Spain b Engineering Projects Department, Universitat Polit ecnica de Val encia, Camino de Vera s/n, 46022, Valencia, Spain c Universidad del Bosque, Av. Cra 9 No.131 A - 02, Bogot a, Columbia article info Article history: Received 14 June 2016 Received in revised form 9 January 2017 Accepted 19 January 2017 Available online 31 January 2017 Keywords: RGB-D sensors Workstation layout Genetic algorithms abstract RGB-D sensors can collect postural data in an automatized way. However, the application of these devices in real work environments requires overcoming problems such as lack of accuracy or body parts' oc- clusion. This work presents the use of RGB-D sensors and genetic algorithms for the optimization of workstation layouts. RGB-D sensors are used to capture workers' movements when they reach objects on workbenches. Collected data are then used to optimize workstation layout by means of genetic algo- rithms considering multiple ergonomic criteria. Results show that typical drawbacks of using RGB-D sensors for body tracking are not a problem for this application, and that the combination with intelli- gent algorithms can automatize the layout design process. The procedure described can be used to automatically suggest new layouts when workers or processes of production change, to adapt layouts to specific workers based on their ways to do the tasks, or to obtain layouts simultaneously optimized for several production processes. © 2017 Elsevier Ltd. All rights reserved. 1. Introduction RGB-D sensors 1 are devices capable of detecting the distance to the objects found in a scene. They are basically composed of an infrared laser transmitter that projects a speckle pattern on its environment, and an infrared camera that captures the projected pattern. The data obtained by the infrared camera is compared with reference standards, allowing estimating the distance of each im- age pixel to the sensor (Garcia and Zalevsky, 2007; Henry et al., 2012; Khoshelham and Elberink, 2012). In addition, the device has a conventional RGB camera, so pixel color data is added to distance data. This technology is not new. There are devices able of capture the depth of a scene from the late twentieth century. However, two features of the new RGB-D sensors have contributed to their increasing use in a wide range of fields and applications. Firstly, the new RGB-D sensors have a very low cost compared to older devices (approximately $ 200 in 2016). Secondly, these new sensors can be used as 3D motion capture systems because the software that controls the sensor provides information about the position of the joints of recognized users present in the scene (skeleton data) in close to real time. Position of body parts are obtained from depth data using a randomized decision forest algorithm, learned from millions of training examples (Han et al., 2013). In 2010 Microsoft released the Kinect RGB-D sensor. Simulta- neously other sensors appeared on the market such as the ASUS Xtion. Since then, the availability in the market of RGB-D sensors fostered the development of promising approaches to usual prob- lems in many areas like object recognition, 3D reconstruction, augmented reality, image processing, robotic, or human-computer interaction. In the field of ergonomics, early research on the uses of the new RGB-D sensors focused on whether the accuracy of the data obtained allowed their use as markerless motion capture de- vices (Bonnech ere et al., 2013a, 2013b; Clark et al., 2012, 2013; Destelle et al., 2014; Dutta, 2012; Fern andez-Baena et al., 2012; Nixon et al., 2013; Obdrz alek et al., 2012; Pfister et al., 2014). The results of these studies suggest that the accuracy of the sensors is only slightly lower than that of more expensive devices, but reliable enough to be used for postural assessment. Other studies focused * Corresponding author. E-mail addresses: jodiemas@dpi.upv.es (J.A. Diego-Mas), ropobau@dpi.upv.es (R. Poveda-Bautista), dgarzonl@unbosque.edu.co (D. Garzon-Leal). 1 Depth sensors are known by several names (Flash Lidar, time-of-flight camera, ranging camera, range sensor …). In this work we will use the term RGB-D sensors because this is currently the most common denomination for this kind of devices. Contents lists available at ScienceDirect Applied Ergonomics journal homepage: www.elsevier.com/locate/apergo http://dx.doi.org/10.1016/j.apergo.2017.01.012 0003-6870/© 2017 Elsevier Ltd. All rights reserved. Applied Ergonomics 65 (2017) 530e540