外文翻译原文-动态虚拟和真实机器人对感知的安全等待时间和机械手最大工作半径的影响.pdf

International Journal of Production Research Vol. 50, No. 1, 1 January 2012, 161176 Impact of dynamic virtual and real robots on perceived safe waiting time and maximum reach of robot arms Parry P.W. Nga, Vincent G. Duffybcd*and Gulcin Yucelbe aSchool of Industrial Engineering and Engineering Management, The Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong SAR, China; bSchool of Industrial Engineering, Purdue University, West Lafayette, Indiana, USA; cRegenstrief Center for Healthcare Engineering, Purdue University, West Lafayette, Indiana, USA;dSchool of Agricultural and Biological Engineering, Purdue University, West Lafayette, Indiana, USA; eSchool of Industrial Engineering, Istanbul Technical University, Istanbul, Turkey (Final version received February 2011) This research examines perception of dynamic objects and robots in a virtual and real industrial work environment. The studies are modelled after those of Karwowski and Rahimi from the early 1990s. By applying virtual reality technology, the real workplace can be simulated in the virtual world for the improvement of facility design. Perception of hazard and risk, safe waiting time, maximum reach of robot arm are measured related to the impact of parameters such as robot size, speed and type and exposure to a virtual accident. Analysis includes techniques such as sequential experiments to compare results in the virtual and real environments. These methods may be considered as a model for studying perception and transfer in other domains. The comparison of the analysed data in the virtual and real environments helps to further determine the transferability of performance and perception from virtual reality to real. Results show similarity in perceived safe waiting time, but there are large differences in perceived maximum reach of robot arms between the virtual and real environments. Using the preliminary results from the integrated data in the sequential experiments, potential guidelines for using virtual facility layout in industry are discussed. Keywords: humanrobot interaction; perception of hazard and risk; maximum reach of robot arms; sequential experiment; data bridging 1. Introduction Many industrial companies utilise industrial robots to perform dangerous tasks in industry to avoid possible hazards. Still, others are attempting to reduce musculoskeletal disorders through the use of hybrid automation such as assist devices (Nussbaum 2000). According to the World Robotics 2007 published by Union Nations Economic Commission from Europe (), there were approximately 951,000 units in 2006 and it is expected to be 1,200,000 units in 2010 (Table 1). Since the robots are used more frequently in workplaces, issues related to humanrobot interaction (HRI) are more often considered by researchers and practitioners. According to Dhillon et al.s (2002) survey, 523 papers about robot safety and reliability were published between 1973 and 2000. Most of this research was done between 1982 and 1990, and robot safety and reliability research has been incrementally decreased since 1986 (Dhillon et al. 2002). On the other hand, between 1995 and 2000, multidisciplinary research in HRI has been more recently started by the collaboration of researchers from human factors, robotics, cognitive science, psychology and natural language, and during that time many conferences and workshops were dedicated to HRI such as IEEE International Symposium on Robot and Human Interactive Communication (RoMan), Association for the Advancement of Artificial Intelligences (AAAI) Symposia Series, IEEE International Conference on Robotics and Automation (ICRA) (Goodrich 2007). Since 2006, international HRI conference has been held annually. One of the issues related to HRI is the safety of the users. In reference to the statistics on Occupational Injuries compiled by the Labour Department of Hong Kong Government in 2007, there were 3967 injuries and 21 fatalities in the manufacturing industry in total. They indicated that there had been 491 workers injured due to striking against or being struck by dynamic moving objects (the number one type of accident) in the manufacturing industry (Hong Kong Labour Department 2007). Previous research on robot safety has been done along two separate *Corresponding author. Email: ISSN 00207543 print/ISSN 1366588X online ? 2012 Taylor one included experiments both with simulated and physical robots. Because of cost and reliability concerns, most of the time it is not possible to conduct experiments with real robots. On the other hand, in simulation experiments, the real worlds detail situations cannot be represented very well. Therefore, in this study the experiments were done both in real and virtual environments. Analytical results are compared in the dynamic virtual and real environments with moving hazard using the Table 1. Summary of number of robots working in industry worldwide. Number of robots Working in industry worldwide951,000 Projected to be working by 20101,200,000 Per 10,000 manufacturing employees in Japan349 Per 10,000 manufacturing employees in Republic of Korea187 Per 10,000 manufacturing employees in Germany186 Per 10,000 manufacturing employees in the United States99 Source: World Robotics (2007), United Nations Economic Commission for Europe. 162P.P.W. Ng et al. sequential experimental techniques (Snow and Williges 1998). Based upon the results obtained from the experiment, the transferability of the experience, perception from virtual to real worlds and similarities and differences of the results given by Karwowski (Karwowski et al. 1988a, 1988b, Karwowski and Pongpatanasuegsa 1990, Rahimi and Karwowski 1990, Karwowski and Rahimi 1991) and this research can be shown. 1.1 Safe robot speed The causes of accidents related to robots could be ascribed to some human perceptual, physical and psychological limitations including human perception of robot size, speed and range of motion that can affect the human behaviour (Carlsson 1984). Different speed of robots can cause different perceptions of hazards. Kulic and Croft (2006) and Ikuta et al. (2003) used velocity as an input while developing a danger index during HRI. The force exerted by robot arms is high with fast speed of robot motion. However, it should be noted that Haddadin et al. (2007) conducted crash tests with robot and dummy head to decide the impacts of collisions between robot and human. They reported that a robot, with arbitrary mass driving moving at speeds up to 2m/s cannot be dangerous to a non-clamped head with respect to the severity indices used in the automobile industry that are based on head acceleration. It can also be noted that other research reported that the human was not in danger for impact with the human chest, abdomen and shoulder at robot velocities up to 2.7m/s. Beside robot velocity, robot masss affect on head injury criterion (HIC) was also investigated and it was reported that a heavy robot cannot pose a significant threat to the human head by means of HIC (Haddadin et al. 2008, 2009). Even though, the safe robot operating conditions (such as speed under 2.7m/s, mechanical output under 150N, etc.) remove physical risks, HRI still involves risks related to the mental strains caused by robot motion (Aria et al. 2010). It should be emphasised that this study is focused on cognitive aspects of robot safety. In this study, robot speeds are chosen 25 and 90cm/s for experiments, above and below the thresholds of concern, since it was previously shown that people feel threatened by robot speed above 64cm/s (Karwowski and Rahimi 1991). Aria et al. (2010) studied mental strains of a human operator in a cell production system where an operator assembles a product with the aid of parts feeding by the robot. Based on their physiological assessment and subjective assessment results, the operator feels discomfort when the robots speed is more than 500mm/s. It can also be mentioned that the initial impact may not be the greatest reason for the concern expressed by the operators. Especially with large robots, operators are aware that the potential for a pin of body parts against other objects after impact is highly likely if a collision occurs since the robot does not necessarily stop after impact whereas in an auto and in transportation, there is nothing to continue to drive the collided objects together after initial impact. Hence, operators perception of their own reaction time may be influencing their perceived safe robot speed rather than simply a concern over the damage at initial impact. 1.2 Perception of safe robot idle time The American National Safety Standard, American National Standards Institute (1986) ANSI R15.06 was established for robot safety in the United States. Also, the Occupational Safety and Health Administration (OSHA) in the US provided guidelines for robot safety (OSHA 1987). The standards related to robotic safety are summarised in Table 2. According to Bonney and Yong (1985) and Nagamachi (1986, 1988), the complex robot systems are potentially hazardous even in the normal mode of operation. Most accidents happened because robot operators misperceive the reasons for pauses, which are either system malfunctions or programmed stops (Sugimoto and Kawaguchi 1983). Accident reports have shown that people can be injured or be killed by robot arms if they misperceive the work envelope and enter it during the robot operation. 1.3 Simulated accident A simulated accident can be introduced to influence the behaviour since the expected shift in the processing of information brings the task into the cognitive realm (Lehto and Papastavrou 1993, Park 1997). Rahimi and Karwoski (1990) suggested that the idle times must be considered in designing the facility layout and robot programmes. It is expected that the exposure to a simulated accident will influence the waiting time to enter the International Journal of Production Research163 work envelope for both robots. As suggested by Parsons (1986, 1987), it has been shown that the simulated robot accidents influence the robot operator after training (Karwowski et al. 1991). Hypothesis 1 It is expected that factors of exposure to a simulated accident, size, speed and type of robots will affect waiting times (i.e. idle times) significantly in both the virtual and real environments. 1.4 Perception of maximum reach of robot arms The robot work envelope is defined as the maximum reach of robot arms or the unsafe zone of a robot. According to Karwowski (1991), the maximum reach of robot arms were significantly affected by factors such as accident exposure, size, speeds and type of robots. The methodology from Rahimi and Karwowski (1990) and Karwowski et al. (1991) are replicated in this study and the results will be compared. Wright (1995) reported that real world distance perceptions are usually 8790% of actual distances. Lampton et al. (1995) showed the tendency for underestimating distance in both the virtual and real environments, but the distance in virtual were more extremely underestimated than that in real. Witmer and Kline (1998) attempted to determine how accurately stationary observers could estimate distances to objects (i.e. cylinders) in a simple virtual environment, given by static cues for distance and defined perceived distance judgement by referring to tasks in which stationary observers judge the distance between themselves and a stationary or moving object immediately perceivable to them. Based on the results from Witmer and Kline (1998), people generally underestimated distance to the objects in the virtual and real environments, but the errors in distance estimation was to be greater in virtual than that in real. Moreover, they showed that the size of the object (i.e. cylinders) influenced the estimated distance significantly but floor texture and pattern did not. Hypothesis 2 Perception of the work envelope of the robot is related to exposure to a simulated accident, speed, types and sizes of robots in both the virtual and real industrial work environments. 1.5 Sequential experimentation and data bridging A sequential experimentation research strategy was proposed by Williges and Williges (1989), which could be utilised for human factors studies to investigate and examine a large number of independent variables using a series of small sequential studies. The results from the sequential studies can be integrated to build the empirical models to explore the effects of different independent variables and predict human performance (Han 1991). Table 2. Robot safety standards. ANSI/RIA R15.06-199 The American National Safety Standard Robot safety Includes risk assessment, methodology and guidelines for safeguarding robotic system CSA 2434:2003Canadian Standards AssociationSimilar to US standards by minor differences ISO 12100International Standard Office Safety of Machinery Standard Basic concepts, general principles for design ISO 10218International Standard Office Robots for industrial environments safety requirements Requirements and guidelines for the inherent safe design, protective measures, and infor- mation for use of industrial robots. IEC61508Functional safety of electrical/electronic/ programmable(E/E/PE) electronic safety-related systems Requirements to minimise dangerous failures in E/E/PE safety-related systems. OSHAOccupational Safety and Health Administration An interpretation of ANSI standards and a directive concerning of robotics safety Source: Spada (2005). 164P.P.W. Ng et al. Data bridging can be treated as a statistical method for integrating results across sequential studies (Han 1991). If there are no significant differences in the responses from the common data points, the data can be considered as from the same experiment and combined into a common data set in order to build the model (Snow and Williges 1998). Based on these results, it is believed that a comparison of virtual and real experiments could be allowed if the data could be merged into a common data set based on the use of data bridging in the virtual experimental conditions. 2. Methods 2.1 Subjects for robot experiment Sixty-four (32 males and 32 females) engineering students were recruited from the Hong Kong University of Science and Technology (HKUST). The subjects of the experiment had a basic understanding about robot programming and operations. The experiments took about 2h. Each participant was paid 200 Hong Kong dollars (7.8HKD1USD) for their participation. All participants were divided into eight groups with eight participants in each group. 2.2 Equipments (robot experiment) Two industrial robots (Yaskawa MOTOMAN-K10S and SONY SRX-410) were investigated in this research. Both robots are located in the CAD/CAM laboratory of the Hong Kong University of Science and Technology. Yaskawa MOTOMAN-K10S is a vertically articulated robot with six degrees of freedom and is mounted on the floor. Its controller is a servo-drive controlling system. The payload capacity of the robot is 10kg. The position repeatability of the robot is 0.1mm. The base rotation of the robot is 320 degrees about the base. The maximum reach of robot arm of MOTOMAN-K10S is 1555mm, and the combined linear speed of all axes is 1500mm/s. Its position repeatability is 0.1mm. The SONY SRX-410 is a SCARA-type high-speed assembly robot. It is a compact desktop design with four axes DC servo motor control. The work envelope of the robot is 600mm (first arm: 350mm; second arm: 250mm). The maximum speed of linear motion (first and second arms combined) is 5200mm/s. The weight of the robot is 60kg (132.2lbs). Its payload capacities are 5kg (at low speed), 3kg (at medium speed) and 2kg (at high speed). The position repeatability of robot for the X/Y-axis and Z axis are 0.025mm and 0.02mm, respectively. Figures 1 and 2 show the Yaskawa MOTOMAN-K10S and SONY SRX-410 robot, respectively. The real workplace with two robots is simulated in the dynamic virtual world by using the Virtual Reality Modeling Language (VRML). A Pentium III (600MHz) computer with a Sony 1700Video Display Terminal (VDT) monitor (V-frequency 75Hz; H-frequency 60kHz) was used and Java Script provided animation in the VRML environment. In the VRML environment, the robot motion was controlled (start and stop) through the program (buttons of control panels) using an Internet Explorer browser with Cosmo player plug-in. Since the VRML program is a virtual internet-based system, the computer is needed to connect the network to the Internet World Wide Web (WWW). Figures 3 and 4 show the MOTOMAN K10S and Sony SRX-410 robots in the virtual environment, respectively. 2.3 Experimental design In this study, the four experiments can be divided into two main categories: (1) idle time experiment and (2) maximum reach of robot arm experiment. Each category of the experiment had two parts: one was the virtual part and the other one was the real part. All participants belonging to the real group were required to perform some virtual trials to fulfil the requirement of data bridging for the sequential experiment in order to do the common point testing. For the real group, half of the subjects (16 subjects) performed tasks in the real environment first and the other half performed some tasks in the virtual environment first. A simulated accident was shown to half of the subjects (16 subjects in the real group and 16 subjects in the virtual group). International Journal of Production Research165 2.3.1 Idle time experiment This experiment was a six-factor ANOVA mixed design (2 genders?2 accident exposures?2 speeds?2 sizes? 2 types of robots?2 lighting levels). All independent variables had two levels. The between-subject variables are gender (males or females) and accident exposure (Yes or No). The other four within-subject independent variables are speed of the