Differences in Fitts Law Task Performance Based on Environment Scaling Gregory S. Lee and Bhavani Thuraisingham Department of Computer Science University of Texas at Dallas 800 West Campbell Road Richardson, TX 75080 USA leegs@utdallas.edu, bhavani.thuraisingham@utdallas.edu http://www.utdallas.edu Abstract. Haptics research has begun implementing haptic feedback in tasks of great precision and skill, such as robotic surgery. Haptic displays can represent task environments with arbitrary scaling. Fitts Law suggests differences in the scale of a workspace rendered on a visual display and in a haptic display should not affect performance of those tasks. However, interactions of great precision and skill may require understanding and verifying the influence of perceiving an environment when the visual and haptic displays represent those environments with differing scales. This experiment measured the influence that mismatched haptic and visual display scalings had on movement times in. Each of five treatments used different scales in the visual and the haptic displays. A Friedman rank test showed a significant difference across all treatments. A post hoc pairwise comparison showed a nearly significant difference between two treatments. These findings suggest the need for further study using more participants and parametric statistics to measure the magnitude of the possible influences. Keywords: Haptics, Scaling, Fitts Law, Falcon. 1 Introduction Haptic displays provide information to the user through the sense of touch. The advancement of the field has led commonly available high fidelity haptic displays. This availability has also allowed researchers to implement haptics in specialized environments. With the increased complexity and precision of the environments in which haptic displays are employed, care must be taken in understanding what factors may influence these tasks. Research has shown haptic feedback enhances human-computer interaction. For example, it has been shown that humans process information presented through the sense of touch more quickly than through the sense of sight[1]. Also, Hasser et al. demonstrated giving icons haptic forces increased user performance on a point-and-click task[2]. Other work has extended these findings to small icon forces and also shown that increasing the small forces increased performance[3,4,5]. More complex interactions have been studied. Maneewarn and Hannaford have shown that haptics can assist novices in operating a robotic manipulator to operate the M. Ferre (Ed.): EuroHaptics 2008, LNCS 5024, pp. 295 300, 2008. c Springer-Verlag Berlin Heidelberg 2008
296 G.S. Lee and B. Thuraisingham manipulator in kinematically well conditioned configurations[6]. Haptic feedback has been applied to surgical robots to show that surgeons benefit from the feedback[7]. Traditional Minimally Invasive Surgery (MIS) already scales the visual workspace via the video representation of the workspace, however, the surgeon still manipulates the tissue directly. Because of the abstraction that surgical robotics allows, scaling the interaction has been suggested[8,9]. This enhancement can be applied to many tasks. This research used Fitts Law to motivate a means to measure impact on performance scaling a scene differently in the haptic and visual displays may have. 2 Methods To perform the experiment, participants controlled a small cursor on the desktop simulation and selected targets with their right hand using the Falcon TM haptic display from Novint Technologies Incorporated. (See Figure 1.) The haptic display fed back force when the cursor traversed icons in the desktop simulation. There were nine icons on the desktop possessing haptic forces that applied 250 millinewtons directed toward the icon center when the cursor was in the outer half of the icon radius. (See Figure 2.) Once participants correctly selected the current green target icon, it turned yellow again and another icon became the current target icon. After every ten movements, the target icon turned red instead of green to indicate that the movement time was not being recorded. (See Figure 2b.) The red icon was always the upper left icon, though the upper left icon could also be the green target icon and not indicate a pause. Participants used the time as needed, but could also choose to continue the experiment without pausing. This process repeated to collect 400 movement times, 80 for each ID. Fitts Law motivated the hypothesis that visual and haptic display mismatches do not effect movement times. It states: ( ) 2A MT = a + b log 2. (1) W Fig. 1. These experiments used The Falcon TM from Novint Technologies Incorporated, a three degree of freedom haptic display capable of nine Newtons of force and possesses a workspace of ten centimeters by ten centimeters by ten centimeters
Differences in Fitts Law Task Performance Based on Environment Scaling 297 2 2 0 0 2 2 2 0 2 (a) The green target icon indicates movement time is being recorded. 2 0 2 (b) The red target icon indicates movement time is not being recorded. Fig. 2. The desktop simulation indicated the current target for the participant to select by being either red or green. Each pulled the end effector towards the center with 250 millinewtons. This equation relates movement time (MT) to the distance between two targets (A) and the size of those targets (W). (See Figure 3.) Linearly scaling an environment does not change the ID. The ID is defined as: ( ) 2A ID = log 2, (2) W MT = a + b (ID). (3) The experiment consisted of six sessions, the first of which trained subjects in the experiment protocol and allowed participants to become familiar with the device and the task. Training also minimized learning effects. Breaks of a minimum of one minute and as long as a participant desired separated experiment sessions. The five remaining sessions consisted of a single treatment of the experiment. Each treatment used different scaling factors for the environment displayed visually and haptically. (Table 1.) The experiment used three possible lattice spacings; one centimeter, two centimeters and four centimeters. The icon radius was one quarter the lattice spacing. This yielded the same five IDs for every representation. The experiment assigned the order of the treatments randomly to further minimize learning effects. The training session was identical to treatment 1. Fig. 3. A two target Fitts Law task similar to the original Fitts Law experiments[10]. Participants rapidly alternate tapping one target then the other.
298 G.S. Lee and B. Thuraisingham Table 1. Lattice grid spacing for the five experiment treatments. The icon had a radius of one quarter the grid size. (The scaling ratio is not reduced.) Treatment Visual Display Haptic Display Scaling Number Spacing Spacing Ratio 1 2cm 2cm 1:1 2 2cm 1cm 1:0.5 3 2cm 4cm 1:2 4 1cm 1cm 0.5:0.5 5 4cm 4cm 2:2 Five men and six women participated in the experiment. Participants included students, staff and faculty of the University of Texas at Dallas. No participants indicated or exhibited any condition which would interfere with the task. No participants had ever used a haptic display such as the Falcon TM or had ever participated in haptics research. Participants received $5US for their participation. The Institutional Review Board (IRB) at the University of Texas at Dallas approved this experiment. Participants also read and signed an Informed Consent Form approved by the IRB which informed them of their rights as participants. 3 Results The MTs and IDs were fitted to equation 3 using a bisquare weighted least squares algorithm and mean movement times (MTs) for each ID calculated. The bisquare algorithm reduces the effect of outliers on the fit by minimizing their weight in the calculation. Data from all subjects are plotted for each treatment in Figure 4. A Friedman rank test performed on the data determined if there were any significant differences in the rank order of the MTs for the same IDs over all treatments. This analysis treated the MT for each ID from a given treatment as a repeated measure. The Bonferroni correction for five repeated measures requires a p-value of less than 0.01 Fitts Law Line Fits Movement Time (MT) 1 0.9 0.8 0.7 Treatment 1 Treatment 2 Treatment 3 Treatment 4 Treatment 5 0.6 0 1 2 3 4 5 Index of Difficulty (ID) Fig. 4. Fitts Law lines fit to experiment data over all subjects per treatment. The gray lines indicate the IDs.
Differences in Fitts Law Task Performance Based on Environment Scaling 299 Table 2. The χ 2 statistics for post hoc pairwise treatment comparisons. Magnitudes greater than 9.94 indicated statistically significant differences including the Bonferroni correction. ( indicates values approaching significance.) Treatment 1 2 3 4 5 1 1 1.24-8.00-4.29-6.67 2 1-9.24-5.53-7.91 3 1 3.71 1.33 4 1-2.38 5 1 to achieve the equivalent relevance of a p-value of 0.05[11]. The Friedman rank test returned an overall p-value of 0.0071, which indicated significant difference in the data. Post hoc pairwise tests compared individual treatments. (See Table 2.) Based on Tukey s method, in combination with the Bonferroni correction, the absolute values in Table 2 must be greater than 9.94 to indicate a significant difference. The magnitude of the statistic representing the difference in rank of treatments 2 and 3 approaches significance. 4 Discussion A Friedman rank test showed a significant difference in the movement times across treatments. Post hoc pair-wise comparisons showed a nearly significant difference between treatments 2 and 3. These treatments showed the visual workspace at the same scales as each other, but represented the haptic workspaces at 1 / 2 and two times the size of the visual representation respectively. The pair-wise comparisons also showed that treatments 4 and 5 did not show a significant difference in rank. Treatments 4 and 5 represent the workspace in the haptic display at the same scales as 2 and 3, respectively, however, they each represent the visual workspace at the same scale as the haptic display. Also, neither showed a significant difference with treatment 1 which also displays the visual and haptic environment at the same scales. This suggests that scaling mismatch may be what affects the MTs. It is also interesting to note that the differences between treatments 1 and 3 are much greater treatments 1 and 2, though the data does not indicate any statistically significant difference. This suggests that the difference may be more of a factor when haptically growing an environment than for haptically shrinking it. These findings suggest a need to study this effect in greater depth. Two more treatments will help fully investigate the possible phenomenon, the scalings of 0.5:2 and 2:1. Further research will also need more participants to allow processing with parametric statistics which will determine the magnitude of any measured differences. Understanding the subtle influences of mismatching visual and haptic stimuli will aid in the development of haptically enabled tasks and haptic interface paradigms that exert less undesirable influence on the performance of a complex task by a skilled practitioner.
300 G.S. Lee and B. Thuraisingham Acknowledgments. The author wishes to thank the Post-Doctoral Program at the University of Texas at Dallas for supporting this research and Professor Blake Hannaford of the University of Washington Department of Electrical Engineering for additional guidance. The author also wishes to thank the reviewers for their careful and insightful suggestions. References 1. Gray, R., Tan, H.Z.: Dynamic and predictive links between touch and vision. In: Experimental Brain Research, pp. 50 55. Springer, Heidelberg (2002) 2. Hasser, C.J., Goldenberg, A.S., Martin, K.M., Rosenberg, L.B.: User performance in a GUI pointing task with a low-cost force-feedback computer mouse. In: Proceedings of the ASME Dynamic Systems and Control Division, vol. 121, pp. 151 156, American Society of Mechanical Engineers (1998) 3. Dosher, J., Lee, G., Hannaford, B.: How low can you go? Detection thresholds for small haptic effects. In: Proceedings USC Workshop on Haptic Interactions, February 2001, Prentice- Hall, Englewood Cliffs (2001) 4. Lee, G.S., Hannaford, B.: Preliminary two dimensional haptic thresholds and task performance enhancements. In: Haptics Symposium, pp. 85 90. IEEE,Los Alamitos (March 2003) 5. Lee, G.S.: Low Power Haptic Devices: Ramifications on Perception and Design. PhD thesis, University of Washington (2004) 6. Maneewarn, T., Hannaford, B.: Haptic feedback of kinematic conditioning for telerobotic applications. In: Proceedings of International Conference on Intelligent Robots and Systems, pp. 1260 1265 (November 1998) 7. Mayer, H., Nagy, I., Knoll, A., Braun, E.U., Bauernschmitt, R., Lange, R.: Haptic feedback in a telepresence system for endoscopic heart surgery. Presence: Teleoperators for Virtual Environments 16(5), 459 470 (2007) 8. Satava, R.M.: How the future of surgery is changing: Robotics, telesurgery, surgical simulators and other advanced technologies. Technical report, University of Washington Medical Center (May 2006) 9. Rosen, J., Hannaford, B.: Doc at a distance. IEEE Spectrum 43(10), 34 39 (2006) 10. Fitts, P.M.: The information capacity of the human motor system in controlling the amplitude of movement. Journal of Experimental Psychology 47, 381 391 (1954) 11. Abdi, H.: The Bonferonni and Šidák corrections for multiple comparisons. Encyclopedia of Measurement and Statistics 12. McSweeney, M.: Nonparametric and Distribution-Free Methods for the Social Sciences. Wadsworth Publishing (1977)