• Federal Highway Administration >
• Publications >
• Research Publications >
• 10025 >
• 03.Cfm >
• Operating Characteristics of the Segway™ Human Transporter

Operating Characteristics of the Segway™ Human Transporter

Chapter 3. Results

The results related to travel speed, approach speed, acceleration, stopping distance (planned and unplanned stops), and clearance distance are presented in the following sections.

Speed

Travel Speed

Table 5 shows participants' observed mean speeds for each key, with the standard error of the mean provided in parentheses. An analysis of travel speed over the midsection of the course indicated that, as expected, participants traveled faster when using faster keys, F (2, 12) = 352.875, p < 0.001. In addition, there was no significant difference in speed whether they were performing a planned or unplanned stop, F (1, 6) = 0.408, p = 0.546. These observed speeds illustrate that riders felt comfortable riding near the maximum speed allowed by each key. Recall that participants had been asked to ride as fast as they felt they could safely ride in each key.

Table 5. Mean observed travel speeds for the three speed keys.

1 mi = 1.61 km Note: Standard error of the mean is provided in parentheses.

Approach Speed

Another operating characteristic of the Segway^TM HT that is of interest to transportation professionals is the approach speed, which was examined for both novice and experienced riders on two different sections of sidewalk. The wide section of sidewalk involved riding past pedestrians and inanimate objects, while the narrow section of the sidewalk involved riding past inanimate objects only. Trials were grouped by sidewalk section type and obstacle type. Participants started each trial from a laterally random starting point and used the yellow speed key during the entire phase II experiment.

There were two major environmental conditions: (1) wide sidewalks and (2) narrow sidewalks. Two separate mixed group analyses of variance (ANOVAs) were conducted for the two different combinations of the variables. Table 6 portrays the results of these global ANOVAs (main effects only).

Table 6. Results of ANOVAs for approach speed.

E = Experience. T = Trial. O = Obstacle.

As can be seen in table 6, three effects were evaluated: (1) the experience of the Segway^TM HT rider, (2) the trial number (an indicator of practice or learning), and (3) the different obstacles in the path. With regard to approach speed on the wide sidewalk, all three independent variables had a statistically significant effect. Only one interaction was statistically significant, which was obstacles by trials (O x T), F (8, 11) = 4.15, p = 0.016, indicating that the nature of the obstacle and trial effects on approach speed may be somewhat different.

In the case of the narrow sidewalk, the outcome was the same as for the wide sidewalk. With regard to approach speed, all three independent variables had a statistically significant effect. Only one interaction was statistically significant, which was the obstacles by experience (O x E), F (3, 108) = 4.88, p = 0.004, indicating that the nature of the obstacle and experience effects on approach speed may be somewhat different.

In summary, the outcomes of the global ANOVAs indicated that the different obstacles in the path had a significant effect on the approach speed of Segway^TM HT riders on both the wide and narrow sidewalk sections. On both types of sidewalk, the experience of the rider and the trial number had a significant effect on the speed with which the rider approached an obstacle.

Obstacle Effect

As can be seen in table 7, for the conditions containing obstacles in the path, the mean overall approach speed was 4.5 mi/h (7.3 km/h) for all participants. In the table, standard errors are shown in parentheses. Across all participants, the mean baseline speed was 1.0 mi/h (1.6 km/h) faster than the mean speed approaching obstacles. This difference in mean approach speed was statistically significant, t (19) = -5.687, p < 0.001. Such a speed differential might be expected between an open pathway with no obstacles and one with obstacles present.

Table 7. Mean approach speed comparing obstacle conditions.

1 mi = 1.61 km Note: Standard errors are provided in parentheses.

Table 7 also reveals that the novice participants approached the obstacles slower than the experienced participants by 1.9 mi/h (3.1 km/h) on average. This difference in mean approach speed was statistically significant, t (18) = -5.993, p < 0.001.

As can be seen in the table, across all participants, the baseline condition revealed a mean speed of 5.5 mi/h (8.8 km/h). In the yellow key mode (medium speed setting), the Segway^TM HT had a maximum speed of 8 mi/h (12.9 km/h). The somewhat lower mean baseline speed was likely a result of the relatively short total length of the sidewalk test sections, which did not allow the participants to accelerate to a maximum travel speed. In phase I, where there was a much longer sidewalk for acceleration, the mean travel speed was 7.7 mi/h (12.4 km/h). Therefore, the results may be more representative of crowded urban conditions than open suburban conditions.

Sidewalk Effect

Table 8 shows the effect of the width of the allowed sidewalk path for all conditions with obstacles present. In the table, standard errors are given in parentheses. For all participants and all obstacles on the wide sidewalk, the mean overall approach speed was 4.8 mi/h (7.8 km/h). On the narrow sidewalk, the participants approached the obstacles slower than on the wide sidewalk by 0.5 mi/h (0.8 km/h) on average, traveling at a mean approach speed of 4.3 mi/h (6.9 km/h). This difference in mean approach speed was statistically significant, t (19) = -2.877, p < 0.001. The slower speed on the narrow sidewalk was probably the result of the constrained navigation space under this condition.

Table 8. Mean approach speed comparing experimental sidewalk sections.

1 mi = 1.61 km Note: Standard errors are provided in parentheses.

Trial Effect

Table 9 shows the effect of the three trials that were used to compute each mean when obstacles were present in the travel path. Standard errors are shown in parentheses. The table reveals the extent of any practice or learning that may have occurred on repeated drives past the same obstacle. For all participants, in general, the mean approach speed increased by 0.3 mi/h (0.6 km/h) over the three trials, indicating that some practice or learning may have taken place. This general increase in mean approach speed was statistically significant, F (2, 38) = 23.597, p < 0.001. The observed increase in approach speed may reflect increased confidence on the part of both groups of participants as a result of repeated practice passing the same obstacle on the sidewalk.

Table 9. Mean approach speed comparing trial effect.

1 mi = 1.61 km Note: Standard errors are provided in parentheses.

Obstacle and Sidewalk Type Comparisons

As can be seen in table 10, for all participants with regard to speed, the baseline condition resulted in the highest mean speeds, passing the cone resulted in the slowest mean speeds, and passing a pedestrian resulted in intermediate mean speeds. In the table, standard errors are given in parentheses. The pedestrian walking toward and pedestrian walking with conditions represent a pedestrian moving toward the participant and a pedestrian moving in the same direction as the participant, respectively. In general, the moving pedestrians were passed by the Segway^TM HT rider at an average speed of 5 mi/h (8.1 km/h). These values were greater than the average approach speed of 4.7 mi/h (7.5 km/h) for inanimate traffic control devices. The differences in speed were statistically significant, F (3, 57) = 5.716, p = 0.002. The 0.3 mi/h (0.5 km/h) faster speed for passing pedestrians as opposed to inanimate objects is such a small speed differential that it is not likely to be of practical consequence.

Table 10. Mean approach speed for obstacles on a wide sidewalk.

1 mi = 1.61 km Note: Standard errors are provided in parentheses.

As can be seen in table 11, with regard to approach speed, for all participants, the baseline and the barrier alone mean approach speeds were approximately the same for the narrow sidewalk. This outcome might be expected, given that the actual sidewalk width was the same throughout for both conditions, though the barrier alone condition had traffic control barriers on either side of the sidewalk approximately 36 inches (914 mm) high. Both barrel conditions resulted in lower mean approach speeds, with the barrel and barrier condition having the lowest mean speed out of all of the obstacle conditions. This outcome might also be expected, given that the most physically constraining condition in the entire experiment was the combined barrel and barrier obstacle on the narrow sidewalk. The standard errors are shown in parentheses.

Table 11. Mean approach speed for obstacles on a narrow sidewalk.

1 mi = 1.61 km Note: Standard errors are provided in parentheses.

Overall Comparisons

Table 12 shows the overall results of the experiment for mean approach speed. Standard errors are given in parentheses. The table reveals the mean approach speed for all of the research participants regarded as a single group as well as for the novice and the experienced participants regarded separately.

Table 12. Mean approach speed overall.

1 mi = 1.61 km Note: Standard errors are provided in parentheses.

Acceleration

Because of the design of the course in phase I, it was possible to gather performance data in both directions-first for the stopping trials and second for the return trip back to the beginning of the course. This second set of data was used to further understand how riders adjust their speed.

Following each stop, participants rode to the end of the course and turned around, beginning their return trip back to the start of the course. Their speed was measured over the marked areas of the course to determine acceleration behavior.

Figure 7 shows the mean speed profile for riders over an approximately 100-ft (30.5-m) length of course. Minimum and maximum speeds are also plotted for each speed key.

1 mi = 1.61 km 1 ft = 0.305 m

Figure 7. Graph. Mean speed during the return trip as a function of distance traveled in each speed key.

Participants took more distance to reach a relatively constant speed when they were using a faster speed key. Note that the wider range between minimum and maximum speeds in the red key compared to the yellow and black keys. These data indicate that all riders felt comfortable enough to ride near the top of the speed range for the black and yellow keys, but some riders did not feel as comfortable riding in the maximum speed range associated with the red key. Several maximum speed values measured for the black, yellow, and red keys were above their respective speed limits of 6 mi/h (9.7 km/h), 8 mi/h (12.8 km/h), and 12.5 mi/h (20.1 km/h). These values may be the result of the Segway^TM HT being able to temporarily exceed the speed limit as well as the result of errors in the speed estimation procedure.

Distance

Stopping Distance

A benefit of having participants travel as fast as possible in a given speed key is that such a procedure allowed for the evaluation of stopping while traveling at high speeds (which should hypothetically result in the longest stops). Stopping performance was analyzed based on the time and distance traveled between the initiation of a stop and the completion of a stop. Both planned stops and unplanned stops were evaluated.

Planned Stops

Braking time and distance are shown in table 13, which shows when the researchers observed that a rider initiated a stop until the stop was complete. Standard errors are shown in parentheses.

Table 13. Mean braking time and mean braking distance for planned stops.

1 mi = 1.61 km 1 ft = 0.305 m Note: Standard errors are provided in parentheses.

As participants approached the predetermined stop location, they made a decision to initiate the stop. Once the researcher observed that they had initiated a stop, it took on average almost 2.5 s and over 10 ft (3.1 m) to complete a planned stop.

Unplanned stops

In addition to braking time and distance, unplanned stops included a measurable response component (see table 14). Response time is the time between the activation of the signal light and the observation that the participant initiated a stop. Response distance is the distance traveled during this time. Standard errors are shown in parentheses.

Table 14. Mean response time and mean response distance for unplanned stops.

1 mi = 1.61 km 1 ft = 0.305 m Note: Standard errors are provided in parentheses.

The mean response time was approximately 0.5 s, and the mean response distance was approximately 5.4 ft (1.6 m). Differences in response time, F (2, 12) = 0.820, p = 0.400, and response distance, F (2, 12) = 3.977, p = 0.089, as a function of mode (key) were not statistically significant. Before riders initiated a stop in the unplanned condition, they had already traveled over 5 ft (1.5 m) and used 0.5 s on average with an SD of 0.2 s. This may be a conservative estimate of real-world performance. For unplanned stops, the participants always expected to see a signal, but they were not sure exactly when it would occur. Response time and distance might be longer under real-world conditions if a rider is operating at a lower level of vigilance and at a higher level of uncertainty as to the type and location of stimuli which might require a stop. Using a different methodology, Landis et al. observed a mean perception-reaction time of 1.1 s with an SD of 0.6 s.^(8,9)

Braking time and distance for unplanned stops, which is the time and distance traveled between the initiation of braking until the completion of the stop, are shown in table 15. Standard errors are shown in parentheses.

Table 15. Mean braking time and mean braking distance for unplanned stops.

1 mi = 1.61 km 1 ft = 0.305 m Note: Standard errors are provided in parentheses.

After the participants began to stop, it took them, on average, almost 2 s and over 9 ft (2.7 m) to complete an unplanned stop. To understand the totality of unplanned stopping behavior, it is important to look at the combination of the response component with completing the stop itself (i.e., total stopping time = response time + braking time). Table 16 shows the mean total stopping time and stopping distance for unplanned stops. Standard errors are shown in parentheses.

Table 16. Mean total stopping time and mean total stopping distance for unplanned stops.

1 mi = 1.61 km 1 ft = 0.305 m Note: Standard errors are provided in parentheses.

On average, unplanned stops took over 2 s and about 14.5 ft (4.42 m) when considering both the response and braking components of the trial.

Comparing Planned and Unplanned Stops

Planned and unplanned stops are intrinsically different because the unplanned condition includes a measurable response component. Table 17 merges the results described above and shows a comparison of performance for these two types of stops. Standard errors are shown in parentheses.

Table 17. Comparisons of performance for planned and unplanned stops.

1 mi = 1.61 km 1 ft = 0.305 m Note: Standard errors are provided in parentheses.

Overall, speed mode had a significant effect on braking distance, F (2, 12) = 219.703, p < 0.001, but not braking time, F (2, 12) = 2.307, p = 0.168. Faster speeds required longer distances to stop regardless of the type of stop being made. Contrasts indicated that mean braking distances differed significantly among all speed keys: black (mean = 5.5 ft (1.7 m), SD = 0.6 ft (0.18 m)), yellow (mean = 8.7 ft (2.6 m), SD = 0.88 ft (0.27 m)), and red (mean = 14.7 ft (4.48), SD = 0.94 ft (0.28 m)). Thus, participants traversed longer distances in approximately the same length of time. There were no other observed significant effects of speed mode (key), stop type (planned versus unplanned), or any interactions.

Clearance Distance

As previously noted, phase II of the experiment investigated the passing behavior of novice and experienced riders on two different sections of sidewalk. The trial conditions were identical to those described in the approach speed section. For clearance distance, two separate mixed group ANOVAs were conducted for the two different combinations of the variables. Table 18 shows the results of these global ANOVAs (main effects only).

Table 18. Results of ANOVAs for clearance distance.

NS = Not significant. E = Experience. T = Trial. O = Obstacle.

As can be seen in the table, three effects were evaluated: (1) the experience of the Segway^TM HT rider, (2) the trial number (an indicator of practice or learning), and (3) the different obstacles in the path. With regard to clearance distance on the wide sidewalk, only the different obstacles in the path had an observed statistically significant effect. In this instance, two interactions were statistically significant: (1) obstacles by trials (O x T), F (8, 144) = 2.89, p = 0.005 and (2) trials by experience (T x E), F (2, 144) = 3.48, p = 0.042. These interactions indicate that the particular effects that obstacles, trials, and experience have on clearance distance may be somewhat different.

In the case of the narrow sidewalk, the outcome was the same as for the wide sidewalk. With regard to clearance distance on the narrow sidewalk, only the different obstacles in the path had a statistically significant effect. In this instance, only one interaction was statistically significant: pass (P) by experience (P x E), F (1, 108) = 6.81, p = 0.018, indicating that the nature of the pass and experience effects on clearance distance may be somewhat different. The independent variable P refers to the outgoing or return pass of the object and is only relevant for the narrow sidewalk where each trial consisted of two determinations. The effect of P was not statistically significant for clearance distance, so it does not appear in the table.

In summary, the outcomes of the global ANOVAs indicated that the different obstacles in the path had a significant effect on the clearance distance of Segway^TM HT riders on both the wide and narrow sidewalk sections. On both types of sidewalk, the experience of the Segway^TM HT rider and the trial number did not have a significant effect on the clearance distance around that obstacle.

Obstacle Effect

As can be seen in table 19 below, across all participants, the baseline condition revealed a mean bias of 0.2 inches (5.1 mm) to the left of the sidewalk center (positive deviations represent the right side of the Segway^TM HT). Without being given instructions to ride on any particular portion of the sidewalk, participants showed a mean travel path almost exactly down the center of the sidewalk, showing only a slight bias toward the left. It should be noted in this regard that baseline distance measurements are relative to the center of the Segway^TM HT. By contrast, obstacle clearance distance measurements are relative to the outboard edge of the nearest wheel of the Segway^TM HT. Since the Segway^TM HT was 25 inches (635 mm) wide, the difference is 12.5 inches (317.5 mm). Standard errors are shown in parentheses.

Table 19. Mean clearance distance comparing baseline versus all obstacle conditions.

1 inch = 25.4 mm Note: Standard errors are provided in parentheses.

As can also be seen in the table, across all conditions containing obstacles in the path, for all participants, the mean overall clearance distance was 14.5 inches (368.3 mm). The experienced participants did not pass the obstacles farther than the novice participants. The difference in mean clearance distance of 1.6 inches (40.6 mm) on average was not statistically significant.

Sidewalk Width Effect

Table 20 shows the effect of the width of the allowed sidewalk path for all conditions with obstacles present. For all participants and all obstacles on the wide sidewalk, the mean overall clearance distance was 25.1 inches (637.5 mm). On the narrow sidewalk, the participants passed the obstacles closer than on the wide sidewalk by 17.7 inches (449.6 mm) on average. This difference in mean clearance distance was statistically significant, t (19) = 17.78, p < 0.001. The closer passing clearance distance on the narrow sidewalk was likely the result of the constrained navigation space under this condition. Standard errors are shown in parentheses.

Table 20. Mean clearance distance comparing experimental sidewalk sections.

1 inch = 25.4 mm Note: Standard errors are provided in parentheses.

Trial Effect

Table 21 shows the effect of the three trials that were used to compute each mean when obstacles were present in the travel path. The table reveals the extent of any practice or learning that may have occurred on repeated drives past the same obstacle. For all participants, in general, the mean clearance distance did not decrease over the three trials. The apparent general decrease in mean clearance distance of about 0.3 inches (7.6 mm) was not statistically significant. Standard errors are shown in parentheses.

Table 21. Mean clearance distance comparing trial effect.

1 inch = 25.4 mm Note: Standard errors are provided in parentheses.

Obstacle and Sidewalk Type Comparisons

As can be seen in table 22, with regard to clearance distance, for all participants, passing a pedestrian walking in the same direction resulted in the greatest mean clearance, and passing the barrel resulted in the lowest mean clearance. There was a difference of 26.1 inches (662.9 mm) between these two mean clearances on the wide sidewalk. In general, the moving pedestrians were passed at an average clearance of 35.9 inches (911.9 mm). The average passing clearance distance for inanimate traffic control devices was 14.2 inches (360.7 mm). Therefore, on average, the extra clearance distance the Segway^TM HT riders afforded to pedestrians was 21.7 inches (551.2 mm). This difference in clearance distance was statistically significant, F (3, 57) = 74.46, p < 0.001. Standard errors are shown in parentheses.

22. Mean clearance distance of obstacles on the wide sidewalk.

1 inch = 25.4 mm Note: Standard errors are provided in parentheses.

As can be seen in table 23 below, with regard to clearance distance, for all participants on the narrow sidewalk, the barrel alone and barrel and barrier conditions resulted in the shortest clearance distances, with the barrel and barrier being the shortest out of all of the conditions. This outcome may be taken as evidence of the relative physical constraints of the narrow sidewalk condition. For all participants, the barrier alone condition resulted in a much higher clearance distance relative to the right wheel of the Segway^TM HT, placing the Segway^TM HT on a path close to the middle of the narrow sidewalk path. Standard errors are shown in parentheses.

Table 23. Mean clearance distance of obstacles on the narrow sidewalk.

1 inch = 25.4 mm Note: Standard errors are provided in parentheses.

Overall Comparisons

Table 24 shows the overall results of the experiment for mean clearance distance. It reveals the mean clearance distance in inches for all of the research participants regarded as a single group as well as for the novice and the experienced participants regarded separately. Standard errors are shown in parentheses.

Table 24. Mean clearance distance overall.

1 inch = 25.4 mm Note: Standard errors are provided in parentheses

Previous | Table of Contents | Next