Autonomous Bus Pilot Project Testing and Demonstration using Light Rail Transit Track

By taking advice from local police and referring to the “Guideline for Pilot Test of Self-driving System on Public Roads (The National Police Agency, 2015)” [7], a safety check presented in Table 2 was conducted to confirm the feasibility of an AB proceeding down the track.

In the final preliminary safety check on the day of the pilot test, we confirmed safe self-driving of ABs based on the conditions presented in the test items in Table 3.

Two cases of AB driving positions relative to the LRT stop shed depicted in Fig. 14 were compared. Consequently, although the AB runs on the bumpy surface of the LRT tracks in Case 1, it did not affect the precise docking control, lane-keeping, and lane changing systems, as shown in Figs. 15 and 16. We rejected an alternative option of a driving position closer to the stop, which would assure more precise docking. The reason being that there should be at least a 483 mm gap from the inner side of the left-hand front wheel of AB to the edge of stop. However, the actual gap with level surface from the bump was only around 400 mm, making this option unsuitable (Fig. 17).

To maintain security during the pilot test, visible plates “Pilot Program of Autonomous Bus in Tram Track @ Hiroshima” are displayed on the front, side, and rear faces of AB to enable general road users to identify the self-driving vehicle (see Figs. 18 and 19). To case of emergencies such as obstacle avoidance, one driver and one operator were ready to force to stop the AB.

During the test, general traffic was restricted not to enter the road section of the pilot test to ensure safety (Figs. 20 and 21). Using safety cones and allocating traffic control staff, the LRT track was enclosed with 2 lanes and a car lane. This ensured that no general cars entered the test section, as shown in Fig. 22.

The time, date, contents and traffic restriction of pilot test was announced to the residents via their neighborhood associations. And those were also publicly announced to journalistic organizations before the test on November 5 (Table 4).

Data including lateral position, distance (velocity by differentiating with respect to time), lateral error, acceleration (traveling direction), and acceleration (lateral) were observed from 9 repetitive test-drives, as shown in Fig. 29. Table 7 shows the used sensors of self-driving controls.

The traveling direction velocities varied during the docking by ACC at the rear of stop and during exiting the LRT track (Fig. 29). We determined that this was due to recognition errors of censors relative to buildings near the LRT stop, traffic security agents, safety cones for traffic restriction, and other factors. The lateral errors tended to have a positive sign, indicating that is the AB most often moved to the left side at stop and were uneven among repetitive drives (Lateral error in Fig. 29).

The traveling direction accelerations varied up to \(\pm\) 0.1G at the departure from a stop during self-driving. On the other hand, lateral accelerations were observed to vary up to \(\pm\) 0.05G at entering/exiting track (see acceleration (traveling direction) and acceleration (lateral) in Fig. 29).

Two-step stop of the AB was autonomously performed by the ACC system before the previous LRT stopping, as shown in Fig. 30. The target distance of the 1st-step stop was set as 6.0 m, and then the AB ran with a slow speed of 20 km/h up to the 2nd stop, which was set to 3.0 m. The errors observed in each of the 9 test drives were considerably different: the 1st-step stop varied from -0.75 m to 2.55 m, and the 2nd-step stop varied from -0.60 m to 2.95 m, as shown in Fig. 30 and Table 8. It appears that these variations were mainly due to identification sensors reacting to surrounding buildings and variation in deceleration by the braking control.

The age distribution of monitors was almost equal proportions among groups except for monitors in their 30 s, as shown in Fig. 32. However, the gender distribution was unbalanced—the sample was 83.3% male, as shown in Fig. 33.

In terms of their current travel behaviors, 93% of monitors usually use the existing bus service. But infrequent users of “once a month” (31%) and “every 2 weeks” (16%) were the majority, while only 6.9% of respondents do not use the bus at all (see Fig. 34). In contrast, the number of LRT users was less than bus users. Around 27% of monitors did not have experience using the LRT (see Fig. 35).

Most monitors are aware that the advantage of AB service is “Reduction of traffic accidents” both before and after the test-drive. Interestingly, 19% of monitors’ attitudes improved after the test drive regarding the advantage of “Reduction of traffic congestion,” as shown in Fig. 36.

In terms of the disadvantage of self-driving technology, the risk of “Unable to flexibly respond to unexpected circumstances” accounted for over 60%. Particularly, it is an intriguing result that risk there was a 12.5% increase on this measure after the test drive, as shown in Fig. 37.

93% of monitors reported their acceptance for the introduction of ABs into society, and this figure increased by 4.2% after the test drive (see Fig. 38), indicating feasibility of AB service.

68% of monitors stated they were willing to use the AB service before the test drive, and which increased by 4.2% following the test drive (see Fig. 39) again, indicating feasibility for the AB service and a benefit from exposure via demonstration.

After the test drive, many monitors stated their positive evaluations of AB driving down LRT track for the items of “Make certain passengers travel with peace of mind”, “Reduce accident risks” and “Make convenient transfers between buses and the LRT” (see Fig. 40). The experienced users tended to have more positive response to the AB policy, suggesting again exposure benefits and the high potential for public acceptance of AB in the future.

Like monitor survey, the resident survey aimed to sample an even distribution of age groups, excluding younger generations people in their 20 s (see Fig. 41). The resident survey sample achieved a more equal gender balance, with almost half the sample being female, as shown in Fig. 42. Thus, the following analysis of this survey will help determine an unbiased estimate of the public acceptance of AB.

The frequency of bus users is presented in order of descending prevalence: 41% reporting “Non-users,” 35% reporting “Once a month,” and 9% reporting “Every 2 weeks,” (see Fig. 43). With regard to the main reasons not to use bus services, the percentage of “Private mode available” was highest at 38%, followed by reasons related to inadequate levels-of-services (LOSs) such as “Longer travel time to destination” at 33%, “Unsuitable schedules” at 30%, and “unsuitable routes” at 26% (see Fig. 44). The ratio of non-bus users in this resident survey was 6.9% higher than that in the monitor survey.

On the other hand, the percentage of non-LRT users was 51%, followed by “Once a month” at 36% (see Fig. 45). And the main reason not to use the LRT was “Unsuitable routes” (see Fig. 46). The ratio of non-LRT users was much higher than that of non-bus users.

Together, these findings suggest that there are a variety of issues facing traditional public transportation modes that could will be solved by the future introduction of fully autonomous vehicles.

Some drive support systems of Level 2 and 3 AVs are already available—namely, the automated braking system, adaptive cruise control system (ACC), and lane-keeping assist system. The history of these advanced systems may affect risk perception of and willingness-to-use ABs. Figure 47 presents the result as of March 2020, in which 16 to 20% of respondents had experience of these systems.

Over 80% of residents were aware of self-driving or autonomous technologies (see Fig. 48). They acknowledged some advantages of self-driving technology such as “Reduce traffic accidents” and “Saving the troubles of manual driving,” (see Fig. 49), while they recognized the disadvantage of “Unable to flexibly respond to unexpected circumstances” because 47% respondents chose this item (see Fig. 50). These figures closely align with those observed in the monitor survey.

70% of monitors agreed with the introduction of AB driving in the LRT track, while 30% disagreed (see Fig. 51). The main reasons for positive responses were “Looks like to reduce traffic accidents,” and those of negative responses were “A fear of accidents by system errors” and “A sense of uncertainty” (see Figs. 52 and 53). Finally, the current public transport users stated their choice intentions between the existing public transport and newly introduced connected AB-to-LRT. The results revealed a virtually equal divide in public opinion toward these transportation technologies (see in Fig. 54). In future research, we will investigate more into public opinions and how they are shaped over time. Based on the other indicators and previous research, it seems that more exposure to and experience with ABs/AVs would improve public opinion toward these technologies.