Development and Evaluation of Connected-Vehicle-Enabled Optimal Dynamic Path Planning with Bus Stops

Abstract

The frequent stops of transit buses significantly block lanes on roads and generate vehicle queues behind. The passenger cars traveling behind buses may be stuck in the queues and miss the green light in the downstream intersection. They will be very tempted to make lane changes to avoid the stopping vehicles efficiently. However, without knowing the information of bus stations and traffic signals, it is very difficult and dangerous for the queued vehicles to make lane changes at last minute. In this paper, an optimal dynamic path planning system will be developed to assist passenger cars avoid buses so as to improve their mobility on local roads. The system utilizes connected vehicles to receive stop information, including times, duration, and locations, of buses, and the signal timing information from intersections. The information is applied to predict the delay of connected vehicles caused by the buses and intersections. The system also estimate optimal paths for the target vehicles to make lane changes and overpass the buses and the downstream intersection to minimize its travel time delay. In this paper, both synthetic and realistic examples are designed with microscopic traffic simulations to evaluate the performance of the proposed system. The results indicate that the travel time delay for connected vehicles can be reduced by up to 35%. In addition, a sensitivity analysis of the market penetration rates of connected vehicles and demand levels is conducted to understand the benefits and reliability of the system under different stages of the connected environment.

Keywords

operations connected evaluation ITS vehicles

In most cities, especially downtown areas with heavy traffic, the frequent stops of transit buses and delivery trucks can disrupt traffic flow and result in severe congestion. For example, to facilitate easy and timely access to getting back onto the route, many bus stops do not have any designated bays at the roadside for picking up and dropping off passengers. Because of that, stopping buses will block lanes, reduce the capacities of roads, and generate vehicle queues and congestion upstream. Moreover, they will also waste green light time at the downstream intersection. Similar to buses, some other commercial vehicles, such as delivery trucks, mail trucks, garbage trucks, and ride-hailing vehicles, also block main traffic lanes or partially block roads when they stop in bike lanes. In addition to the congestion caused by the stopping buses and delivery trucks, the vehicles in the queues behind find it very difficult to make lane changes to avoid the stopping vehicles because of the high speed of traffic in the adjacent lane and the consequent long waiting time and high risk of collision. And, as a result of the delay, they may also miss the green light at the downstream intersection. In that sense, there is a very urgent need to develop an advanced assistance system to improve the mobility of the queued vehicles.

In the literature, the congestion impact of public transits, especially bus stations, on a road segment or network had been widely investigated. The impact of bus stations on traffic flows were well analyzed with both mathematical models (1 –4) and field observations ( 5 ). These studies indicated that the bus frequencies, dwell time, stop locations, stop types, and number of lanes had significant impact on the capacity of roads and the stability of traffic flow. Gu et al. ( 6 ) showed that near-side bus stops which were close to intersections could reduce the delay to passenger cars, while delays to buses could be reduced with far-side bus stops which were far away from the intersections. Shen et al. ( 7 ) demonstrated the advantage of near-side bus stops for improving the urban network while they affected the bus-carrying capacity. In addition, the impact of bus stops on a large-scale network was also evaluated with microscopic traffic simulations ( 8 ). Johari et al. ( 9 ) investigated the impact of bus stop locations on the network macroscopic fundamental diagram. Their study showed that the near-side bus stops reduced network performance, that is, produced smaller capacities and higher network average delays. To reduce the negative impact of bus stops, two major strategies were being implemented. The first strategy focused on optimizing the locations and design of bus stops under different congestion and demand levels (10 –12); the second method relied on transit signal priority to assign green lights for buses to pass intersections without stopping (13 –16). Both methods were able to minimize the delay to buses as well as reduce road congestion caused by buses and their stops. However, they were not designed to assist the queued vehicles to avoid buses, especially in their immediate responses to bus stops.

Even though specific studies for the assistance of the queued vehicles were not well represented, there was a lot of effort spent on solving similar problems, such as road congestion caused by traffic incidents or work zones. In Yamada et al. ( 17 ), Harada et al. ( 18 ), and Xing et al. ( 19 , 20 ), variable message sign (VMS) systems were developed with lane-level traffic information and provided lane-changing instructions to balance traffic on all lanes as well as to mitigate traffic congestion caused by incidents. However, the VMS systems required installing signs at fixed locations on roads, and this was very expensive to implement if there were a lot of buses or if delivery trucks were involved. Recently, the development of connected vehicles brought innovative solutions for the problem with low cost. Connected-vehicle-based variable speed limit (VSL) control was one widely applied strategy to mitigate congestion from vehicle incidents. Most VSL systems utilized PID (proportional-integral-derivative) control ( 21 ) and model predictive control (22 –25) to restrict flow entering the regions of nonrecurrent congestion as well as to maximize the discharge flow rates. These systems were very effective at mitigating road congestion.

Moreover, connected-vehicle-enabled lane-changing assistance systems were also developed to instruct vehicles passing nonrecurrent road congestion caused by traffic incidents. In Moriarty and Langley ( 26 ), several cooperative lane selection strategies were developed to search for the optimal lane for each vehicle to maintain high speed on the road using supervised and reinforcement-learning algorithms. The strategies relied solely on the sensors installed on the vehicles to monitor the surrounding traffic and to make decisions. In Jin et al. ( 27 ), a real-time optimal lane selection (OLS) algorithm was established to reduce vehicle travel times. The algorithm applied connected vehicles to collect location, speed, lane, and desired driving speed information of individual vehicles along a road, and a lane selection agent determined the optimal lane for each vehicle to improve the system-wide benefits such as travel time and fuel consumption. In Ye and Ramezani ( 28 ), a lane distribution optimization was proposed for autonomous vehicles to choose the optimal lanes to reduce freeway congestion. Xiaoping et al. ( 29 ) utilized connected vehicles to coordinate lane-changing behaviors of vehicles near nonrecurrent freeway bottlenecks, and the system estimated advisory lane instructions for connected vehicles to make early lane changes to pass bottlenecks. In Kang et al. ( 30 ), the authors developed an optimal lane-changing advisory system using connected vehicles. For each connected vehicle, the system defined a speed utility for each lane based on the downstream connected-vehicle dynamic information, and advised the vehicle to make lane changes to the optimal lane with the highest speed utility. Yang and Oguchi ( 31 ) extended the work in Kang et al. ( 30 ) and developed an advisory lane-changing assistance system to help connected vehicles bypass road incidents. The system estimated the optimal lane for each connected vehicle based on the information collected by all connected vehicles, and it was able to shorten the congestion region caused by road incidents. However, the authors of that study did not work on solving the specific problem of queued vehicles behind stopped buses or delivery trucks.

One pioneer approach to the problem was proposed by Ratnasingam ( 32 ), who invented the idea of determining the lane instructions for vehicles to avoid lanes with frequent stops of buses, delivery trucks, garbage trucks, and so forth. However, the idea was not evaluated with any experiments, and the system did not consider the impact of traffic signals. Moreover, it was not able to provide an immediate response for queued vehicles to avoid bus stops. In this paper, an optimal dynamic path planning (ODPP) system will be developed with the help of connected vehicles to overcome this challenge. The system uses connected vehicles to share information on vehicle dynamics and the signal timing plan among buses, signals, and passenger cars. A delay prediction model is proposed to estimate the extra travel time delay to upstream vehicles caused by bus stops at intersections. In addition, the connected vehicles behind buses will be provided with lane-changing instructions based on the predicted delay to minimize their travel time. The system is also evaluated with both synthetic and real-world experiments using microscopic traffic simulations to understand its mobility benefits for both connected and non-connected vehicles, and a sensitivity analysis of market penetration rates (MPRs) of connected vehicles and traffic congestion levels is conducted to understand the reliability of the system under different traffic conditions and at different stages of connected-vehicle development.

The rest of the paper is organized as follows. The next section describes a delay prediction model and the development of the ODPP system with the help of connected vehicles to improve the mobility of queued vehicles caused by buses (and other commercial vehicles). The third section evaluates the proposed system with both synthetic examples and microscopic traffic simulations, and a sensitivity analysis of MPRs of connected vehicles is also conducted. Finally, the fourth section concludes the findings and makes some recommendations.

Optimal Dynamic Path Planning System

To assist queued vehicles in the upstream of a bus, it is very important to understand the dynamics of the bus and the traffic in the upstream. In this section, a delay prediction model of the queued vehicles is proposed based on vehicle-to-vehicle (V2V) and vehicle-to-signal (V2S) communications, followed by the development of the ODPP system.

Delay Prediction of Queued Vehicles

In this subsection, a delay prediction model is developed to calculate the extra travel time delay caused by the bus stops. Figure 1 shows the queue generated by a single bus on a road, where there is a bus stop ahead of the intersection. In the system, we assume that there are several connected vehicles and one bus ahead of the intersection. Both the connected vehicles and the bus are equipped with wireless communication devices, and they are able to share their dynamics, including location, speed, and lane information, with each other. Moreover, they are able to communicate with the signal to obtain the signal phasing and timing information.

Figure 1.

Vehicle queue behind a single bus.

Assume that the intersection is the starting point, and the bus stop is located at distance $x_{S}$ upstream of the intersection (see Figure 2). Also assume that all connected vehicles and buses can communicate with each other and the signal when they are within distance of $L$ , that is, the control region, upstream of the intersection. In addition, at time $t$ , the bus is traveling ahead of the intersection, and its location and speed at time $t$ are set as ${x_{b} (t), v_{b} (t)}$ , respectively. Also, it is traveling on the rightmost lane in the control region to access the stop. The bus can broadcast its dynamics and the stop information, including whether it will stop at the bus stop $S_{b} (t)$ and the dwell time $δ_{b}$ , to all connected vehicles. Moreover, assume that there are $N (t)$ connected vehicles in the rightmost lane behind the bus, and we denote the location and speed of the connected vehicle $k$ as ${x_{k} (t), v_{k} (t)}, k = 1, 2, \dots, N (t)$ , respectively. Meanwhile, the connected vehicles will talk to the intersection to obtain the signal phasing and timing plan, that is, the start time $g_{s}$ and the duration $δ_{g}$ of the upcoming green light and the duration of the next red light. $δ_{r}$ .

Figure 2.

Demonstration of vehicle information with bus.

Based on the speed and location of the bus at time $t$ , its anticipated arrival time at the bus stop is $t_{a} = \frac{x_{b} (t) - x_{S}}{v_{b} (t)}$ . If the bus does not stop at the bus stop, that is, $S_{b} (t) = 0$ , then there is no extra delay caused by the bus. If the bus broadcasts to the following connected vehicles that it will stop at the bus stop, the extra delay to the connected vehicle $k$ will be estimated. The dynamics of both bus and regular passenger cars can be analyzed under four different scenarios as shown in Figure 3.

Figure 3.

Dynamics of bus and passenger cars ahead of intersection: (a) no delay, (b) partially block I, (c) partially block II, and (d) fully block.

Figure 3a shows the first scenario, no delay. In this scenario, if the bus does not stop at the bus stop, it will hit the queue generated ahead of the intersection by the red light. If the bus stops at the bus stop for the duration of $δ_{b}$ , it still enters the same queue, and the following connected vehicle (the green line) will also be delayed at both the bus stop and the intersection. In that sense, the extra delay caused to the connected vehicle by the bus stop will be $0$ , as the connected vehicle will still be queued ahead of the intersection even without the bus stop.

In the second scenario (see Figure 3b), partially block I, if the bus does not stop at the bus stop, it is anticipated to pass the intersection without hitting the queue ahead of the signal. However, if it stops at the bus stop, it will block the lane for the duration $δ_{b}$ , and the connected vehicle will be queued behind the bus. If the connected vehicle can also pass the intersection in the current green light, the delay is defined as its stopping time in the queue behind the bus. Assume the times and speeds at which the bus and the connected vehicle $k$ enter the control region are ${t_{b, c}, v_{b, c}}$ and ${t_{k, c}, v_{k, c}}$ , respectively. With that, the queue length ahead of the bus can be estimated as $N_{k, b} = \sum_{i = 1}^{k} (n_{i} + 1)$ , where $n_{i}$ is the number of vehicles between connected vehicle $i - 1$ and $i$ , and it can be estimated by

n_{i} = Q (v_{i, c}) \cdot (t_{i, c} - t_{i - 1, c}), \forall i = 1, 2, \dots, k,

(1)

where $Q (\cdot)$ is the flow–speed relationship of the lane from the fundamental diagram. If $i = 1$ , $n_{i}$ indicates the number between the bus and the first connected vehicle, and $t_{0, c} = t_{b, c}$ . Therefore, the delay is expressed as

d_{k} = t_{a} + δ_{b} + \frac{N_{k, b} \cdot L_{v}}{ω} - (t_{k, c} + \frac{L - N_{k, b} \cdot L_{v}}{v_{k, c}}),

(2)

where $L_{v}$ is the jam spacing (including the average length of passenger cars and the minimum clearance between two consecutive vehicles), and $ω$ is the queue-releasing speed on the lane. The value of $ω$ is determined by the capacity of the lane $q_{c}$ , the density at capacity $ρ_{c}$ , and the jam/maximum density $ρ_{j}$ , that is, $ω = \frac{q_{c}}{ρ_{j} - ρ_{c}}$ .

Figure 3c shows the third scenario, partially block II, where the bus can pass the intersection during the green light if it does not stop at the bus stop. If it stops at the bus stop, it can still pass the intersection during the same green light, but the connected vehicles will miss the green light and wait ahead of the intersection. Then, the delay of the connected vehicle $k$ can be estimated under two situations.

The first situation occurs if the connected vehicle $k$ can pass the intersection at the green light when the bus does not stop at the bus stop, that is, $t_{k, c} + \frac{L}{v_{k, c}} \leq g_{s} + δ_{g}$ . Based on the arrival time of the vehicle in the control region, the queue length ahead of the signal can be estimated as

N_{k, s} = N_{k, b} - \frac{v_{f}}{ω + v_{f}} \cdot ω (t_{a} + δ_{b} - (g_{s} + δ_{r}) - \frac{x_{S}}{v_{f}}),

(3)

where $v_{f}$ is the road speed limit. Then, the extra delay caused by the bus stop can be expressed as

d_{k} = (g_{s} + δ_{g} + δ_{r} + \frac{N_{k, s} \cdot L_{v}}{ω} + \frac{N_{k, s} \cdot L_{v}}{v_{f}}) - (t_{k, c} + \frac{L}{v_{k, c}}),

(4)

where the two component on the right side are the times of the vehicle passing the intersection when the bus skips and stops at the bus stop, respectively.

The second situation occurs if the connected vehicle $k$ cannot pass the intersection at the green light when the bus does not stop at the intersection, that is, $t_{k, c} + \frac{L}{v_{k, c}} > g_{s} + δ_{g}$ , so that the vehicle has to stop ahead of the intersection to wait for the next green light. Assume that the connected vehicle $k'$ can pass the intersection, while $k' + 1$ cannot, that is, $t_{k', c} + \frac{L}{v_{k', c}} \leq g_{s} + δ_{g} < t_{k' + 1, c} + \frac{L}{v_{k' + 1, c}}$ . In that sense, all vehicles behind the connected vehicle $k' + 1$ and some vehicles between $k'$ and $k' + 1$ cannot pass the intersection. With that, the queue length ahead of the intersection before the vehicle $k$ is estimated as

N_{k, s}^{'} = \sum_{i = k' + 2}^{k} n_{i} + \frac{(t_{k', c} + \frac{L}{v_{k', c}}) - (g_{s} + δ_{g})}{t_{k' + 1, c} - t_{k', c}} \cdot n_{k' + 1} .

(5)

Then, the extra delay of the connected vehicle $k$ is

\begin{matrix} d_{k} = & (g_{s} + δ_{g} + δ_{r} + \frac{N_{k, s} \cdot L_{v}}{ω} + \frac{N_{k, s} \cdot L_{v}}{v_{f}}) \\ - (g_{s} + δ_{g} + δ_{r} + \frac{N_{k', s} \cdot L_{v}}{ω} + \frac{N_{k', s} \cdot L_{v}}{v_{f}}) \\ = & (N_{k, s} - N_{k, s}^{'}) L_{v} \cdot (\frac{1}{ω} + \frac{1}{v_{f}}) \end{matrix} .

(6)

Figure 3d demonstrates the last scenario, fully block, where the bus can pass the intersection in the current green light if it does not stop at the bus stop. However, if it stops at the bus stop, it will miss the green light and stop ahead of the intersection for the next green light. The queue length ahead of the intersection is also $N_{k, b}$ for the connected vehicle $k$ , and its anticipated time to pass the intersection is

t_{k, s} = g_{s} + δ_{g} + δ_{r} + (N_{k, b} \cdot L_{v} + L_{b}) (\frac{1}{ω} + \frac{1}{v_{f}}),

(7)

where $L_{b}$ is the jam spacing for the bus. Similar to the third scenario, the estimation of extra delay for connected vehicle $k$ can be categorized into two situations. In the first situation, the vehicle is not queued ahead of the intersection when the bus does not stop at the bus stop. Then, the extra delay caused by the bus stop is

d_{k} = t_{k, s} - (t_{k, c} + \frac{L}{v_{k, c}}),

(8)

In the second situation, the vehicle will be queued ahead of the intersection if the bus stops at the bus stop, and the queue length is $N'_{k, s}$ in Equation 5. Then, the extra delay can be estimated with Equation 9.

d_{k} = (N_{k, s} L_{v} + L_{b} - N_{k, s}^{'} L_{v}) \cdot (\frac{1}{ω} + \frac{1}{v_{f}})

(9)

With all the four scenarios described above, the delay of any connected vehicle behind the bus can be estimated based on their dynamics, the bus stop information, and the signal timing information. The estimated delay will be applied for the development of the ODPP system in the rest of this paper.

System Development

With the extra delay to connected vehicles caused by bus stops, an ODPP system will be developed in this subsection. The system is designed to reduce or remove the delay applied to the connected vehicles. Figure 4 shows the flow chart for the development of the system.

Figure 4.

Flow chart of the ODPP system.

The system utilizes connected vehicles to obtain information on the bus and the signal. The bus will share its location $x_{b} (t)$ , speed $v_{b} (t)$ , and stopping information ( $S_{b}$ and $δ_{b}$ ), and the signal timing plan, including the start time of the green light $g_{s}$ and the duration of green ( $δ_{g}$ ) and red ( $δ_{r}$ ) lights. Meanwhile, each connected vehicle will share its time $t_{k, c}$ and speed $v_{k, c}$ when entering the control region. The delay of each connected vehicle will then be estimated based on the model described above. If there is no extra delay or if the delay is very small (less than the penalty of the lane changes, which is set as 2 s to represent the additional time for lane changes), no instructions will be provided to the connected vehicle, and it will just follow the bus to pass the intersection. If the delay is large, the optimal path will be estimated for the connected vehicle. When the vehicle enters the control region, it will first receive a lane-changing instruction to the adjacent lane so that it can avoid the downstream bus. As shown in Figure 1, the vehicle will move in the adjacent lane at a higher speed and change back to the rightmost lane when it passes the bus if it wants to make a right turn at the intersection. With the optimal path estimation, the connected vehicle can prevent the delay impact from the bus and the upcoming traffic signal.

Results and Discussion

In this section, the proposed ODPP system will be evaluated with both synthetic and realistic networks to understand its benefit for reducing travel delay caused by bus stops. Aimsun, a microscopic traffic simulator, is utilized to model the dynamics of all vehicles and the entirety of the traffic in both networks. Sensitivity analysis of MPRs of connected vehicles and traffic demand is conducted to understand the reliability and robustness of the system under different stages of connected-vehicle development and traffic conditions.

Case Study I: Synthetic Example

In this subsection, a corridor with three intersections is simulated to evaluate the ODPP system. Figure 5 shows the geometry of the network, and the gap between any two consecutive intersections is set as 500 m. All roads have two lanes. Three bus stops are set ahead of the intersections, and their distances to the intersections are shown in Figure 5.

Figure 5.

Geometry of a corridor.

In the experiment, one bus line will be set from point 1 to 2, and the bus is assigned at every 3 min. The bus will stop at all bus stops, and the loading time at each one is set as 30 s. The simulation will run for 1 h, and 1,750 non-connected vehicles are loaded from 1 to 2 in the simulation. In addition, the connected vehicles are assigned from point 1 to points 3, 4, and 5. Each destination will have connected-vehicle demand set at 75 vehicles per hour. Moreover, all intersections are signalized, and the cycle length and the green time of the through movement are set as 80 s and 40 s, respectively. The offsets of all signals are set as 0 s. In the deployment of the ODPP system, the penalty of each lane change for a connected vehicle is set as 2 s, and the length of the control region is 300 m.

Figure 6 shows the trajectories of the bus, non-connected vehicles, and connected vehicles. Figure 6a indicates that as a result of bus stops, a long queue will be formed behind the bus, and both connected and non-connected vehicles are delayed for a long time to pass the intersection. In the figure, the delays of the three connected vehicles are all over 75 s. However, with the deployment of the ODPP system, the connected vehicles are instructed to make lane changes to avoid the bus stops, and they can catch the green lights on the road (see Figure 6b). With that, their delays can be reduced significantly. For the first two connected vehicles, their delays are less than 25 s, that is, the reduction is more than 64%. For the third one, even it is not blocked by the bus, it will still meet the red light. Therefore, its delay keeps at the same level, and the reduction is less than 5%.

Figure 6.

Synthetic example 1: vehicle trajectories: (a) base case and (b) control case.

In addition to the benefits to individual vehicles, the ODPP system can also improve the mobility of all connected vehicles. By comparing the trajectories of all connected vehicles in the whole simulation, the delay reduction of connected vehicles is as high as 20%. However, because of the intensive lane changes of connected vehicles, the performance of non-connected vehicles, which are all the through vehicles, is worse. The average delay of non-connected vehicles increases by about 8%.

Case Study II: Real-World Network Example

In this subsection, the ODPP system is implemented on a small-scale real-world network, the California connected-vehicle test bed (see Figure 7). The test bed is developed along El Camino Real in Mountain View, California, with a length of 2 mi and a speed limit of 35 mph. It consists of 16 signalized intersections, and they are all installed with roadside units based on DSRC (Dedicated Short-Range Communication) technology, which can broadcast signal phasing and timing information to all connected vehicles. Moreover, there are two major and four minor bus lines scheduled on the test bed. The major lines assign buses at every 12 to 15 min, and the minor lines have an interval of 15 to 20 min.

Figure 7.

California connected-vehicle test bed.

In the Aimsun simulation, more than 5,300 vehicles are uploaded to the network within 1 h, and demand is calibrated with the real-world traffic during morning peak hours. Moreover, 48 buses are scheduled along the four predefined routes and the signal timing plans are optimized with the calibrated traffic conditions. Furthermore, only a portion of the passenger cars are assumed to be connected vehicles, and the MPR varies in different simulation scenarios. The connected vehicles can communicate with both buses and traffic signals in real time to deploy the ODPP system.

In the first experiment, the MPR of connected vehicles is set as 10%, and their demand levels ${100 %, 200 %, 300 %}$ are simulated (the demand level is set with respect to the base case of 5,300 vehicles, with 200% representing twice the simulated vehicle number). Table 1 summarizes the comparison results under different demand levels. They indicate that the proposed system can always reduce the travel time delay of connected vehicles at different levels of demands. At the 100% demand level, the reduction can be as high as 26%, while with higher demand levels, that is when the road is more congested, the reduction is smaller. This is caused by long queues ahead of the intersections at high demand levels. Because of the heavy traffic on the road, it is also difficult for the connected vehicles to make lane changes. In that sense, the connected vehicles have a higher probability of being queued behind the buses, which will reduce the benefit of the proposed system. Meanwhile, the system can reduce the travel time delay for non-connected vehicles at the 100% demand level. However, at higher demand levels, their delay can even increase as the lane changes of connected vehicles further increase the congestion on the road.

Table 1.

Comparison of Travel Time Delays under Different Demand Levels

Demand level (%)		Control
	Base	Connected vehicle		Non-connected vehicle
	Delay (s)	Delay (s)	Diff (%)	Delay (s)	Diff (%)
100	151.8	112.2	−26.1	137.8	−9.2
200	412.3	390.2	−5.4	476.0	15.5
400	559.1	515.1	−7.9	631.9	13.0

Note: diff = difference.

In the second experiment, the network is simulated with 100% demand level, and MPR varies from 4% to 50%. Figure 8 shows the savings in travel time delay for connected and non-connected vehicles under different MPRs. For connected vehicles, with higher MPRs, the savings of travel time delay will be smaller. At MPR = 4%, the delay is reduced by up to 34% and the saving is maintained at the high level when MPR ≤ 30%. However, when MPR > 30%, the saving will be reduced dramatically, and it can be negative when MPR ≥ 50%. This trend can be explained by the distributed control of the ODPP system, where each connected vehicle estimates its own optimal path without coordination with the other connected vehicles. With a smaller number of connected vehicles, it will be easier for them to utilize the high speed of the adjacent lane to reduce the delay. In addition, the system can improve the mobility of the non-connected vehicles. Figure 8 shows that there exists a concave relationship between the delay savings and MPRs. The optimal saving is achieved at MPR = 30%, and the reduction is more than 20%. The mobility benefit to the non-connected vehicles can be explained by the improved traffic conditions arising from connected vehicles and shorter queue lengths. However, with high MPRs, the delay savings can also be negative because of worse traffic congestion caused by a high number of lane changes from connected vehicles.

Figure 8.

Savings in travel time delay under different MPRs.

In a word, both experiments illustrate the mobility benefits of the proposed ODPP system under different demand levels and MPRs of connected vehicles. However, because of the distributed control mechanism of the ODPP system, the system does not work well under high demand levels and high MPRs. This issue will be overcome with centralized control in the future.

Conclusion

This paper developed an ODPP system with the help of connected vehicles to assist passenger cars in passing buses and intersections with smaller delays. The system connected signals, buses, and passenger cars to obtain signal timing information, the dynamics of buses and connected vehicles, and the bus stop information. A delay prediction model was developed with the collected information to predict the extra delay of passenger cars caused by bus stops. The proposed system estimated lane-changing instructions with the predicted delay to assist the connected vehicles in passing stopping buses and intersections faster. The system was also evaluated with both synthetic and realistic examples in microscopic traffic simulations. The synthetic experiment showed that the delays to connected vehicles were reduced by up to 64%, and the system could also reduce the delay of connected vehicles by up to 34% and non-connected vehicles by 21% under a real-world network. Moreover, the impact of demand levels and MPRs of connected vehicles was analyzed. The results indicated that the benefits of the proposed system were reduced when the road was more congested. Also, higher MPRs might result in lower savings in travel time delay for connected vehicles. And, because of the distributed control mechanism, a concave relationship between the delay savings of non-connected vehicles and MPRs was observed, and the optimal MPR was estimated as 30%.

Based on the findings from the study, we have the following recommendations for future studies. First, the system has limited benefit at high MPRs of connected vehicles and high demand levels. It lacks the coordination of connected vehicles in all lanes to achieve higher delay savings at higher MPRs and positive savings for both connected and non-connected vehicles at high demand levels. Second, in this paper, the dwell times of buses at bus stops were assumed to be constant. However, they were highly related to the arrival times and the number of waiting passengers. In the future, a probability model will be applied to describe the dwell times for the estimate of travel time delay for connected vehicles. Third, the system is only evaluated in a small-scale network with a limited number of buses and passenger cars. It would be more promising to implement the system in a large-scale network, especially in downtown areas, with more bus routes, so as to better understand its mobility benefits. Fourth, the current system does not consider the situation that the bus will make a left turn at the intersection after the bus stop, which may potentially block the movements of the connected vehicles. The system will be extended to more general scenarios with different bus routes to avoid such conflicts. Moreover, in this paper, we only evaluated the system on roads with a fixed-time signal plan. The benefits will be more significant if the system can be combined with advanced signal control systems to optimize the paths of both passenger cars and buses. Finally, a field experiment will be conducted using a connected-vehicle test bed to evaluate the real-world benefits of the proposed system.

Footnotes

Author Contributions

The authors confirm contribution to the paper as follows: study conception and design: Hao Yang and Kentaro Oguchi; data collection: Hao Yang and Kentaro Oguchi; analysis and interpretation of results: Hao Yang and Kentaro Oguchi; draft manuscript preparation: Hao Yang and Kentaro Oguchi. All authors reviewed the results and approved the final version of the manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Hao Yang

Kentaro Oguchi

References

Zhao

X.-M.

Gao

Z.-Y.

Jia

The Capacity Drop Caused by the Combined Effect of the Intersection and the Bus Stop in a CA Model. Physica A: Statistical Mechanics and its Applications, Vol. 385, No. 2, 2007, pp. 645–658.

Yuan

Y.-M.

Jiang

Q.-S.

Wang

Trafﬁc Behavior in a Two-Lane System Consisting of a Mixture of Buses and Cars. International Journal of Modern Physics C, Vol. 18, No. 12, 2007, pp. 1925–1938.

Tang

T.-Q.

Huang

H.-J.

The Effects of Bus Stop on Trafﬁc Flow. International Journal of Modern Physics C, Vol. 20, No. 6, 2009, pp. 941–952.

Cassidy

M. J.

Gayah

V. V.

Ouyang

Mitigating Negative Impacts of Near-Side Bus Stops on Cars. Transportation Research Part B: Methodological, Vol. 47, 2013, pp. 42–56.

Kwami

A. V.

Kuan

Y. X.

Zhi

Effect of Bus Bays on Capacity of Curb Lanes. Journal of American Science, Vol. 5, No. 2, 2009, pp. 107–118.

Gayah

V. V.

Cassidy

M. J.

Saade

On the Impacts of Bus Stops near Signalized Intersections: Models of Car and Bus Delays. Transportation Research Part B: Methodological, Vol. 68, 2014, pp. 123–140.

Shen

Cheng

Capacity Approximations for Near-and Far-Side Bus Stops in Dedicated Bus Lanes. Transportation Research Part B: Methodological, Vol. 125, 2019, pp. 94–120.

Nguyen-Phuoc

D. Q.

Currie

De Gruyter

Kim

Young

Modelling the Net Trafﬁc Congestion Impact of Bus Operations in Melbourne. Transportation Research Part A: Policy and Practice, Vol. 117, 2018, pp. 1–12.

Johari

Keyvan-Ekbatani

Ngoduy

Impacts of Bus Stop Location and Berth Number on Urban Network Trafﬁc Performance. IET Intelligent Transport Systems, Vol. 14, No. 12, 2020, pp. 1546–1554.

10.

Ibeas

Á.

Dell’Olio

Alonso

Sainz

Optimizing Bus Stop Spacing in Urban Areas. Transportation Research Part E: Logistics and Transportation Review, Vol. 46, No. 3, 2010, pp. 446–458.

11.

Alonso

Moura

J. L.

Dell’Olio

Ibeas

Á.

Bus Stop Location under Different Levels of Network Congestion and Elastic Demand. Transport, Vol. 26, No. 2, 2011, pp. 141–148.

12.

Tirachini

The Economics and Engineering of Bus Stops: Spacing, Design and Congestion. Transportation Research Part A: Policy and Practice, Vol. 59, 2014, pp. 37–57.

13.

Ngan

Sayed

Abdelfatah

Impacts of Various Parameters on Transit Signal Priority Effectiveness. Journal of Public Transportation, Vol. 7, No. 3, 2004, pp. 71–93.

14.

Yin

Zhang

W.-B.

Zhou

Nakamura

Modeling and Implementation of Adaptive Transit Signal Priority on Actuated Control Systems. Computer-Aided Civil and Infrastructure Engineering, Vol. 26, No. 4, 2011, pp. 270–284.

15.

Zeng

Zhang

Balke

K. N.

Yin

A Real-Time Transit Signal Priority Control Model Considering Stochastic Bus Arrival Time. IEEE Transactions on Intelligent Transportation Systems, Vol. 15, No. 4, 2014, pp. 1657–1666.

16.

Park

B. B.

Lee

Y.-J.

Coordinated Transit Signal Priority Supporting Transit Progression under Connected Vehicle Technology. Transportation Research Part C: Emerging Technologies, Vol. 55, 2015, pp. 393–408.

17.

Yamada

Ouchi

Hirasawa

Yamada

Katayama

Horiguchi

Oikawa

Applicability of AHS Service for Trafﬁc Congestion in Sag Sections. Proc., 12th World Congress on Intelligent Transport Systems, San Francisco, CA, 2005.

18.

Harada

Fukase

Maejima

Xing

Seko

A Study on a Measure against Motorway Trafﬁc Congestion by Balancing Lane Utilization of Trafﬁc. Infrastructure Planning Review, Vol. 26, 2009, pp. 881–888.

19.

Xing

Sagae

Muramatsu

Balance Lane Use of Trafﬁc to Mitigate Motorway Trafﬁc Congestion with Roadside Variable Message Signs. Proc., ITS World Congress, Busan, South Korea, Vol. 15, 2010.

20.

Xing

Muramatsu

Harayama

Balance Lane Use with VMS to Mitigate Motorway Trafﬁc Congestion. International Journal of Intelligent Transportation Systems Research, Vol. 12, No. 1, 2014, pp. 26–35.

21.

Lee

Park

B. B.

Evaluation of Variable Speed Limit under Connected Vehicle Environment. Proc., 2013 International Conference on Connected Vehicles and Expo (ICCVE), Las Vegas, NV, IEEE, New York, 2013, pp. 966–967.

22.

Kattan

Khondaker

Derushkina

Poosarla

A Probe-Based Variable Speed Limit System. Journal of Intelligent Transportation Systems, Vol. 19, No. 4, 2015, pp. 339–354.

23.

Luo

Hadiuzzaman

Fang

Qiu

T. Z.

Assessing the Mobility Beneﬁts of Proactive Optimal Variable Speed Limit Control during Recurrent and Non-Recurrent Congestion. Canadian Journal of Civil Engineering, Vol. 42, No. 7, 2015, pp. 477–489.

24.

Chen

Ahn

Variable Speed Limit Control for Severe Non-Recurrent Freeway Bottlenecks. Transportation Research Part C: Emerging Technologies, Vol. 51, 2015, pp. 210–230.

25.

Farrag

El-Hansali

M. Y.

Yasar

Shakshuki

E. M.

Simulation-Based Evaluation of Using Variable Speed Limit in Trafﬁc Incidents. Procedia Computer Science, Vol. 175, 2020, pp. 340–348.

26.

Moriarty

D. E.

Langley

Learning Cooperative Lane Selection Strategies for Highways. AAAI/IAAI, Vol. 1998, 1998, pp. 684–691.

27.

Jin

Boriboonsomsin

Barth

Improving Trafﬁc Operations Using Real-Time Optimal Lane Selection with Connected Vehicle Technology. Proc., 2014 IEEE Intelligent Vehicles Symposium, Dearborn, MI, IEEE, New York, 2014, pp. 70–75.

28.

Ramezani

Lane Distribution Optimisation of Autonomous Vehicles for Highway Congestion Control. In Proceedings 97th Transportation Research Board Annual Meeting. 18-00346, 2018.

29.

Xiaoping

Dongxin

Shen

Qiqige

Wenbo

Coordinated Control Algorithm at Non-Recurrent Freeway Bottlenecks for Intelligent and Connected Vehicles. IEEE Access, Vol. 8, 2020, pp. 51621–51633.

30.

Kang

Elbery

Rakha

Bichiou

Yang

Optimal Lane Advisory on Freeways within a Connected Vehicle Environment. Proc., 2018 IEEE Intelligent Transportation System Symposium, Maui, HI, IEEE, New York, 2018.

31.

Yang

Oguchi

Advisory Lane-Changing Assistant at Vehicle Incidents with Connected Vehicles. Transportation Research Record: Journal of the Transportation Research Board, 2019. 2673: 595–609.

32.

Ratnasingam

Traffic Aware Lane Determination for Human Driver and Autonomous Vehicle Driving System. US Patent 9,672,734, 2017.