Simulation-based Analysis for Reducing Traffic Congestion in Real Traffic Networks by Demand Spreading over Time

Abstract

The concept of demand spreading over time has been proven to eliminate traffic congestion for simple networks having one link and a single bottleneck. However, no method has been developed yet to implement this concept in real traffic networks. In this paper, the issues that arise when demand spreading over time is applied to real traffic networks are identified as the first step toward realizing its practical use. A method of solving the identified issues was examined, and its effect was evaluated using a traffic simulation. We found that the method is effective in reducing traffic jams and economic losses.

Appendix 1

In subsection 2.3, a solution for the four issues is examined. The method that can be realized by an in-vehicle system is proposed. A flow chart for the proposed method is shown in Fig. 8.

Fig. 8

Flow chart for proposed method

• Issue 1: There exist multiple bottlenecks on each route.

  [Proposal] When there exist multiple bottlenecks on the route, there is no way to judge what congestion (or travel) time that each bottleneck accounts for, if it is calculated for the entire route at once. For that reason, the travel time or congestion time, from the origin to the destination, should not be calculated for the entire route at once.

  Rather, the existence/non-existence of congestion should be determined for each link that is passed, and then the overall congestion (or travel) time is summed together.

• Issue 2: The velocity is not constant.

  [Proposal] The velocity at which a vehicle moves in the upstream part of each link is supposed to be nearly equal to its velocity when the demand spreading over time is applied—i.e., the velocity of the non-congested region. Here, the velocity in the upstream part of each link is supposed to be the velocity of the non-congested region, although the non-congested region does not always exist necessarily during all times. “The travel time excluding congestion time f” is calculated using the velocity at which a vehicle runs in the upstream part of each link. The waiting time w ⁱ caused by the traffic congestion in link i is calculated using Eq. A1

  $$ {w^i} = T_r^i - \left( {\frac{{{L^i}}}{{{V^i}}} + T_{{sw}}^i} \right) $$

  (A1)

  where

  ${T_r}^i$::

  travel time for link i

  V ⁱ::

  velocity at which a vehicle is running, upstream of link i

  L ⁱ::

  length of link i

  $T_{{sw}}^i$::

  expected signal wait time of link i.

  Here, the expected signal for the wait time $T_{{sw}}^i$ is calculated with Eq. 4.

• Issue 3: The reason for any delay is difficult to be distinguished between traffic signals and jams.

  [Proposal] The waiting time ${w^i}$ of each bottleneck link, which was judged by the solution of Issue 1, is calculated. As shown in Appendix-Fig. 8, if ${w^i}$ is more than the signal cycle time$T_{{sc}}^i$, it is judged that “the vehicle was held up by the traffic congestion” and the values of ${w^i}$ are added.

• Issue 4: There is a difference in the travel time according to the turn direction.

  [Proposal] Especially when a right turn has to be made at an intersection, the travel time varies according to the number of oncoming vehicles. However, the number of oncoming vehicles cannot be measured because it is assumed that this method is realized by in-vehicle systems. If, therefore, a vehicle turns right at the next intersection and the velocity upstream of the link is low, the vehicle stops not only at the red light but also at the green light. Even though right-turn timing should be calculated based on the flow rate of the oncoming vehicles, in this method, it is assumed that the neighboring links are congested if the link (in which the calculation target vehicle exists) is congested. Then, the expected signal wait time is judged from the velocity for the link in which a vehicle is present. In particular, it is assumed that more than half of the free-flow velocities belong to the free-flow situations and less than half of the free-flow velocities belong to traffic jam situations. If a vehicle turns right and the velocity ${V^i}$ is less than half of the free-flow velocity $V_f^i$, the expected signal wait time $T_{{sw}}^i$ can be calculated using Eq. A2, instead of Eq. 4. Here, ${V^i}$ is the velocity at which a vehicle travels upstream of link $i$.

  $$ T_{{sw}}^i = {{{\left( {\frac{{(T_{{sr}}^i + T_{{sg}}^i) \times (T_{{sr}}^i + T_{{sg}}^i)}}{2}} \right)}} \left/ {{T_{{sc}}^i}} \right.} $$

  (A2)

  where

  $T_{{sg}}^i$::

  signal time of green light downstream of link $i$.

The method proposed above can be realized with an in-vehicle system that is not limited to a particular area or region, because the method can be realized by obtaining the velocity at the upstream part of a link. Although traffic signal data currently cannot be obtained by an in-vehicle system, it is thought that these data could be estimated by accumulating probe data.

As explained above, we proposed the method of the demand spreading over time considering the following four points: (1) the existence of multiple bottlenecks on each route, (2) the difference in the velocity of vehicles, (3) the need to distinguish between traffic signals and traffic jams, and (4) the difference in the travel time caused by the turn direction.