Motivation and Objectives of the Research

Group-riding is a unique behavioral phenomenon of mixed traffic at intersections. It is frequently observed that motorcycles move in groups while crossing intersections because they would feel safer and more confident in making decision. As the results, there are many interactions between groups of mixed vehicles, so-called inter-group interactions, inside intersection areas. Such interactions have critically influenced the mixed traffic performance as serious conflicts between groups have often caused unexpected congestions and accidents. Moreover, the increasing number of automobiles would have significant effects on the interactions and traffic performance as well. Therefore, this research is aimed to investigate "local traffic" rules that are governing mixed traffic at intersections, to find out behavioral mechanisms of the interactions, and to predict the effects of car share increase in the future. Based on the predicted car share effects, the policy objective of the research is to propose and discuss strategic management concepts for the mixed traffic in transitional periods prior to car-dominated traffic era.

Case Study

Hanoi City (Vietnam) is selected as a case study since motorcycles are the dominant mode in the city and management of mixed traffic in the City is also very critical. Video surveys were conducted at 2 signalized and 2 non-signalized intersections, which are typical in the City, in order to collect traffic date for the analyses. To extract data from the video clips, the computer software, named Vehicle Movement Tracking (VMT) is also developed in the research.

Research Methodology

To archive the research objectives, several behavioral analyses are carried out. Firstly, a gap acceptance behavior model (Model 1) is developed for analyzing gap decisions made by groups of left-turn vehicles at intersections. To construct the model, two assumptions are made. First, the vehicle leading the group can be treated as a group representative decision-maker because once he/she has decided the way to move, the others often follow him/her decision. Second, the group leader is assumed not to change lane. Then, a conventional gap acceptance model is applicable with some modifications. Among many factors, expected waiting time and the number of motorcycles in the left-turn group are especially taken into account. Scheme of the group based gap acceptance behavior is shown in Figure 1.

Secondly, a model of gap negotiations between left-turn and straight-go groups is developed (Model 2). Since the assumption of "no lane change" in Model 1 is relaxed, Model 2 can explain the real interactions better. In an interaction, the left-turn group leader may choose one from the three strategies: "stop/give way", "completely cross", and "incompletely cross". In response to the left-turn leader's decision, the straight-go groups can choose one from the four strategies: "drive on", "run in front", "cut tail", and "stop/give way". An outcome of the interaction can be one of the four strategy combinations: SC1 {left-turn: stop, straight-go: drive on}, SC2 {left-turn: incompletely cross, straight-go: run in front}, SC3 {left-turn: completely cross, straight-go: cut tail}, and SC4 {left-turn: completely cross, straight-go: stop}. Thus, the gap decision made by the left-turn leader is conditional on the straight-go leader's decision. If the straight-go leader does not change lane, the interaction become the gap acceptance behavior as described in Model 1. If the straight-go leader changes lane, several fuzzy logics developed are used to explain the reasons why he/she chooses to "run in front" or "cut tail" of the straight-go group.

Thirdly, once the straight-go leaders have decided to "run in front" or "cut tail", how do they change lateral positions overtime and under different conditions? To answer the question, a straight-go vehicle lateral maneuver behavior model (Model 3) is developed. Conceptually, lateral deviations made by a straight-go leader is assumed to be a function of determinant factors, such as, factors related to vehicle compositions and sizes of the two groups, group leaders' types, and the longitudinal movements of the two leaders (see Figure 2). Straight-go leaders' trajectories are then tracked and re-plotted in a two-dimensional plane in order to analyze relationships between lateral deviation and other factors.

Fourthly, my one theory, named Theory of Piggyback and Mirror, is developed based on the most interesting findings found in the first three behavioral analyses. The Theory is centered on the effects of group-riding phenomenon and the mechanism by which the group-riding effects take place. It is assumed that the other vehicle in the group also generate some "power", termed as "piggyback mass", imposing on the leader of the opposite group. Due to the piggyback mass, the opposite leader might be forced to change behavior as to decelerate or change lane. The subject leader may immediately realize this behavioral change because he used to be in the shoes of the opposite leader many times. Then, he might become more confident as to accelerate and cross the intersection. The behavioral change of the opposite leader is named as "Mirror of Behavior" because the subject leader looks into the mirror and changes his behavior upon the situation. A generalized mechanism of the inter-group interactions is shown in Figure 3. In conclusion, the Theory is aimed to realize and re-explain the effects of group-riding phenomenon and to investigate how a group of vehicles is subjectively defined drivers in the opposite group. These understandings are very valuable for planning and management of mixed traffic, especially at intersections.

Findings

Model 1: Gap acceptance behavior of the left-turn group

Amazingly, it has been found that the number of motorcycles in the left-turn group would make the leader more confident as he/she is more likely to accept a short gap if the number increases. As shown in Figure 4, considering the gap acceptance at 50 percentile, single motorcycle, 10-motorcycle, and 20-motorcycle groups would accept gaps of 1.0s, 0.5s, and -0.2s. Probably, the more number of motorcycles the more confident the group leader will be.

Another interesting finding is that "expected waiting time" factor has completely different impacts on motorcyclists and car drivers, as seen in Figure 5. For motorcyclists, the longer the waiting time the lower the probability of gap acceptance. In contrast, for car drivers, the longer they wait the more likelihood that they accept short gaps. The difference can be explained by the fact that the longer expected waiting time would coincide with the longer/bigger straight-go groups. The left-turn leading motorcycles would feel not so confident and decide to wait until others come and finally cross if they feel confident enough. However, the leading auto drivers would already feel confident and become very sensitive to the expected waiting time. These findings are very interesting and can be explained by the Theory of Piggyback and Mirror.

Model 2: Gap negotiation behavior of the straight-go group

The analysis on gap negotiation behavior has shown that the left-turn auto leaders may force the straight-go motorcycles to laterally deviate to the tails of the left-turn groups. As shown in Figure 6, two typical cases are compared, straight-go motorcycles versus left-turn motorcycles in case 1 and straight-go motorcycles versus left-turn autos in case 2. In the latter, percentage of due to the existence of auto leader, strategy combination {left-turn: completely cross, straight-go: cut tail} shares about 50 percent, while the number in the former is just 12 percent. It is meant that the left-turn autos have some advanced "power" compared to the motorcycles.

Model 3: Lateral maneuver behavior of straight-go vehicles

In the third analysis, it has been found that left-turn autos would make straight-go motorcycles perform larger cut-tail deviations than the motorcycles do. However, in the mean time, the higher number of autos behind the straight-go motorcycles would encourage their confidences as the deviations decrease according to the increased straight-go auto shares. These contrary impacts are also considered as group-riding effects.

Theory of Piggyback and Mirror

In the Theory, the gap acceptance behavior comparisons have shown that the "weights" or powers of one passenger car, minibus, and bus are equivalent to 5, 12, and 24 motorcycles, respectively. It is also found that following vehicles that are 25m or longer far from the leader cannot generate piggyback mass because they are not taken into account by the drivers in the opposite group.

Following hypotheses which contribute to the inter-group interaction mechanism are tested and confirmed:

・Hypothesis 1: If left-turn piggyback mass behind the left-turn leader is bigger, the straight-go leader is more likely to decelerate. Thus, there is a significantly negative correlation coefficient between these two factors;

・Hypothesis 2: If the straight-go leader decreases speed due to the left-turn piggyback mass exceeds critical value (Mcr = 6), the left-turn leader would be more confident as to increase speed, and vise versa. Therefore, there is a significantly negative correlation coefficient between these two accelerations;

・Hypothesis 3: If the power ratio between the left-turn group and the straight-go group increases, the straight-go leader would be less confident as to decrease speed. See Figure 8 below.

In conclusion, group leaders who become leader by chance (so-called deciding leaders) may not be confident enough to cross intersection if they move alone or in small groups. However, if there are many vehicles behind, they will be more confident in the battles because of the piggyback mass generated by some of the followers. Therefore, the leader's decision is inherently the group's will. Regardless of leader type, the winner in the game would be the group with significantly higher "power".

Policy implications

The behavioral models and the Theory are applied to forecast behavioral changes under scenarios of different car share. At present (car share < 11%), percentages of "run in front" and "cut tail" are very high, totally 40% and the deviation max are ・3.0m. However, in the near future, keen attention must be paid on the case in which left-turn car share to be more than 50% because "cut tail" probability would be up to 50% with the deviation max of -5.0m. In the future, though "run in front" and "cut tail" would decrease but could be still high (20%). If so, actions should be taken from now on in order to eliminate the dangerous driving behaviors inherited from the "motorcycling culture". Based on the understandings of the interactive mechanisms and the behavioral change predictions, some management concepts are roughly proposed to cope with the mixed traffic in each transitional period. The proposed measures include motorcycle sub-lane system, car setback areas, road design, multi-phased signal systems, and driver education.

Figure 1 Scheme of group-based gap acceptance behavior

Figure 2 Conceptual movements of straight-go vehicles

Figure 3 Generalized mechanism of inter-group interactions

Figure 4 Impact of the number of motorcycles in the left-turn group on the left-turn leader's gap decision

Figure 5 Behavioral difference between motorcyclists and auto drivers in term of sensitivity to expected waiting time

Figure 6 Impact of left-turn auto on straight-go motorcycle's decision

Figure 7 Impacts of left-turn leader type and straight-go car share on lateral deviations of straight-go motorcycles

Figure 8 Straight-go leader acceleration and group power ratio

近年アジア地域の開発途上国の大都市では急速な経済成長が進行し、オートバイや自動車の保有と利用が急激に伸びている。これに道路整備や信号整備が追いつかず、都市内で激しい交通渋滞が発生している状況にある。これだけでなく、運転者のマナー教育やルール設定が十分に行われていないために交通事故が多発するなどの新たな問題を引き起こしている。

論文提出者はまさにこれらの問題に直面しているベトナムハノイ市の出身である。ハノイでは現在オートバイが都市交通の中心であり、その保有が年15%以上の割合で継続的に増加しているが、このモータリゼーションの軸足は徐々に自動車へと移行しつつあり、街路上でオートバイと自動車が高密に混在して走行する状況となっている。有史以来、このような経験をした都市は大変限られており、オートバイはいずれ消えていく交通機関であるとの認識が強く重要な研究課題として位置づけられて来なかったことから、オートバイと自動車の混合交通流に関する研究蓄積はほとんど存在しないのが実情である。ハノイ市では今後20年程度はオートバイが都市の主要な交通機関を占めることが予測されており、両者の関係性に着目した交通流解析を通じて、今後の道路設計や運用に生かしていくことが急務となっていよう。

ハノイ市の交差点での高密な交通流を観測すると、左折しようとするオートバイと自動車が車群を形成し、これに対応して直進車両も車群を形成し、相互の大きさ関係で交通流の状態が決定されるという大変ユニークかつ支配的な現象が見られる。本論文では、この現象を表現する理論を詳細な交通流観測データから構築することを目的としている。

本論文で行われている分析の流れとその重要な成果について解説する。第1章で上記の問題意識を整理し、第2章では関連する論文レビューを行い、第3章で研究の枠組みを提示している。

第4章では、ハノイ市の主要な信号交差点および無信号交差点で実施した高密混合交通流のビデオ撮影調査の概要を述べている。撮影画像から分析対象とする車群の組み合わせを700程度抽出し、各車群に含まれる車両の位置データを0.5秒ごとに詳細にコーディングした。

第5章は左折車群のギャップ(流入しようとする車群間)選択行動特性を分析しており、左折車群の車両台数が増加するとともに選択するギャップ長が小さくなる傾向があることが明らかとなった。このことは車群の先頭車が後方車両の"力"を利用して直進車群に圧力を掛けつつ強引にギャップ選択を行う行為が背後に存在することを示唆するものである。

第6章では左折車群の先頭車がオートバイと自動車の場合で、直進車群と左折車群の相互Negotiationによって最終的に実現する交通流パターンに影響するか分析し、先頭車が自動車の場合に直進車群が左折車群の後方に回り込もうとする危険な行動("cut tail"と称す)が卓越することが明らかとなった。これは直進車が自動車の大きさが与えるある種の"圧力"を嫌ってその影響を最も受けない走行経路を選択した結果であると解釈可能である。さらに,このcut tailと、逆に直進車群が左折車群から遠ざかる方に車線変更する行為("run in front"と称す)の走行軌跡の差異を分析した。その結果、直進車群の先頭車が自動車の場合にはcut tailがほとんど生じず、これとrun in frontのいずれの場合にも側方移動距離が小さくなることが明らかとなった。これはオートバイと自動車の運動性能の違いに依るところが大きいと予想される。

以上の基礎的な観測に基づいて、車群の先頭車が背後の車両群の大きさを直接確認することなく相手車群の挙動から確信的に把握するような行動の存在を確認するに至った。本論文ではこれを"Piggyback and Mirror Theory"と名付け、第7章でその理論化を試みた。背後車両群の各車両が相手車群に与えるインパクトの大きさ(Piggyback Mass: PM)を車種固有の重みと先頭車からの物理距離の関数として表現し、観測データからこれらを同定した。その上で、(1)左折車群のPMが増加すると直進車群の加速度が減少する、(2)直進車群のPMが一定以上の大きさになると左折車群は発車しない、等の仮説を統計的に検定し、観測データからの理論の妥当性を確認した。

第8章では、今後自動車の混入率が増加する場合に想定される車群間の交通流パターンの特性に関して、両車群内の自動車混入率の違いに着目したシナリオ分析を実施し、自動車混入率のレベルに応じて臨機応変に対応する交通管理手法を導入する必要があることを確認した。

第9章は以上をとりまとめた結論である。

以上、本研究は、交差点においてオートバイと自動車で構成される車群の相互挙動を説明する理論を構築した。この成果を実際の交通管理施策に役立てるにはもう少し研究を蓄積する必要があるが、そのための第一歩として大変有益でユニークな研究であると確信している。ベトナムだけでなく、今後発展が予想される開発途上国の大都市でも本論文の成果は参考となるはずであり、世界的なインパクトも大きい。よって本論文は博士(工学)の学位請求論文として合格と認められる。