WO2019101676A1

WO2019101676A1 - Method and device for dynamically controlling a signal light system

Info

Publication number: WO2019101676A1
Application number: PCT/EP2018/081715
Authority: WO
Inventors: Robert Markowski; Robert Oertel; Jan Trumpold
Original assignee: Deutsches Zentrum für Luft- und Raumfahrt e.V.
Priority date: 2017-11-23
Filing date: 2018-11-19
Publication date: 2019-05-31
Also published as: DE102017221011B4; CN111373454A; DE102017221011A1; EP3714445A1

Abstract

The invention relates to a method for dynamically controlling a signal light system (2). Phases (P0, P1, P2, Pi) of the signal light system (2) are controlled by means of a controller (3) on the basis of time lost, said time lost being the total time lost (21) of all of the vehicles located in a detection radius, wherein the total time lost (21) is estimated by means of a predictive device (4) on the basis of current and estimated future vehicle positions (10-x) of the vehicles in the detection radius. The invention additionally relates to a corresponding device (1).

Description

Method and device for dynamically controlling a traffic light system

The invention relates to a method and a device for dynamically controlling a traffic signal system.

Traffic lights are often used to control road traffic nodes. These have as a primary task to keep the waiting and loss times of the road users by a skillful release time distribution low. Often for this purpose in Germany established control procedures such as fixed-time control or

traffic-dependent rule-based procedures based on vehicle time gaps, vehicle requirements or vehicle occupancy levels used. However, there are meanwhile concepts and technical possibilities for the detection of vehicles, which so far are hardly used. These include, for example, floating car data, vehicle-to-X communication or the capture of vehicle trajectories by means of cameras.

In addition, the established methods take into account network-level interactions at most with static approaches. This severely restricts the flexibility of the procedures and queries that deviate from the original planning basis can lead to inefficient tax behavior of the procedures. Runners, so

Traffic lights without a fixed orbital period are usually controlled by rules. Here, however, often only the traffic situation in the currently approved direction is considered, the overall traffic situation at the node is at most on request loops for other phases in combination with a maximum waiting time or requirements by public transport with a. Freewheelers thus achieve greater flexibility than systems with fixed-time control or a master plan, but almost no longer account for interactions at the network level. So they are only for isolated

Nodes suitable.

To account for interactions, it is known to coordinate fixed-time controls along a traffic corridor. Here, the release starts of the successive traffic signal systems are coordinated so that most road users in the coordinated direction can pass through the network section unhindered. Depending on the time of day, different fixed-time programs can be run at the various junctions, which can be coordinated in different ways Directions can result. For all of them, however, there are usually massive restrictions on traffic quality for non-coordinated directions.

To mitigate these restrictions, framework plans can be used instead of fixed schedules. These are characterized by certain core release times, which are matched to each other as in the fixed-time control, that an unobstructed passage of the network section should be possible. In addition, at the beginning and / or end of the core release time, stretch areas are provided where the release can generally be stretched based on rule-based approaches. In this way, a limited traffic dependence can be established. Especially at high loads of the network section but also this approach has the disadvantage that the phases are generally stretched maximum and thus adjusts the behavior of a fixed-time control again. A disadvantage compared to the fixed-time control is additionally in the poor coordination of the nodes with each other, since the sole vote of the core release times for an ideal coordination is not sufficient.

Basically, three methods are known: Independent-decentralized control methods try to detect interactions at the network level on the basis of local measured values. The advantage here is that no communication infrastructure must be set up to neighboring plants or a central processing unit. This is especially important where many traffic lights are not integrated into a communication network and the construction of such a network would be possible only with great effort. However, the disadvantage of the independent-decentralized control method is that only the local data is available. This can, for example, impair the detection of vehicle sparks and cause the traffic light system to react too late. With an extension of the detection radius, however, this problem could be counteracted.

The second method is based on self-organized decentralized systems. Here, adjacent traffic signal systems can communicate with each other and thus, for example, expand their virtual forecast horizon, exchange requirements or otherwise coordinate with each other and thus further optimize their local control. For this, however, a corresponding communication infrastructure must be present or created. The third method is based on centrally organized systems. Here are the

Traffic lights of all controlled nodes connected to each other via a central processing unit. Here, the central processing unit usually carries out an optimization on the basis of the aggregated measured values of the individual nodes, which as a result delivers, for example, direct control commands or master plans for the individual nodes. In addition to the costs of setting up and maintaining the communication infrastructure, the problem here is the increased complexity of the system. This can lead to long delays during communication, which means that the system can only respond to local changes with a delay. In addition, such a system is also difficult to introduce and expand, since individual components are virtually without function in themselves and new components must be incorporated into an existing, complex overall system.

From DE 10 2009 033 431 A1 a method and a device for the dynamic control of a traffic signal system are known, wherein at least one control unit controls phases of the traffic signal system on the basis of at least one loss time of at least one vehicle.

The invention is based on the technical problem, a method and a

To provide a device for dynamically controlling a traffic signal system in which a dynamic control of the phases of the traffic signal system is improved.

The technical problem is solved by a method with the features of claim 1 and a device having the features of claim 8. Advantageous embodiments of the invention will become apparent from the

Dependent claims.

In particular, a method for dynamically controlling a traffic signal system is provided, wherein phases of the traffic signal system are controlled by means of a control on the basis of a loss time, wherein the loss time a

Is the total loss time of all vehicles in a detection radius, and the total loss time is estimated by means of a forecasting device based on current and estimated future vehicle positions of the vehicles in the detection radius. Furthermore, an apparatus for dynamically controlling a traffic signal system is provided, comprising a controller, wherein the controller is designed to control phases of the traffic signal system by means of a control based on a loss time, the control comprising a prediction device, wherein the loss time is a total loss time of all a detection radius is located, and wherein the predicting means is adapted to estimate the total loss time based on current and estimated future vehicle positions in the detection radius.

The basic idea of the invention is to dynamically control a traffic light system at a node on the basis of the total loss time of all vehicles detected in a detection radius. In particular current, but also future loss times arising in the current circulation of the phases of the individual vehicles are taken into account. The respective loss time is estimated based on the current and future vehicle positions in the detection radius.

The advantage of the invention is that an efficient control of nodes, taking into account interactions at the network level, can be realized without static approaches. The real-time-capable, model-based approach estimates the resulting total loss time at the node, and in this way one

enables loss-time-optimal dynamic control without the need for communication with neighboring nodes and a complex communication infrastructure. The device is not by a fixed orbital period

However, by forecasting the resulting loss times, it can nevertheless take into account interactions at the network level. A dynamic control of the traffic signal system is thus substantially improved.

In one embodiment, to estimate the total amount of time lost by the predictor based on the individual vehicle positions, a number of vehicles in a queue and a queue length are estimated. This allows a queue length to be included in the dynamic control. The queue length is always updated when another vehicle is added to the corresponding inflow to the hub. In a further embodiment, it is provided that for estimating the total loss time by means of the forecasting device, a remaining release time of a phase and a release start of a phase following this phase are estimated on the basis of the number of vehicles in the queue. This estimation then forms the basis for the estimation of the respective loss times of the vehicles in the phases following the current phase. The longer the current phase lasts, the longer vehicles that are not allowed to drive, wait, etc.

In a further embodiment, it is further provided that, for estimating the total loss time by means of the forecasting device, a loss time for each vehicle located in the detection radius is estimated on the basis of the remaining release duration, wherein for the loss time of a vehicle a respective loss of waiting time, a

Response loss time and an acceleration loss time is taken into account. The

Waiting Loss Time refers to the time that a vehicle stops at a standstill

Node must wait. The reaction loss time is the time that the vehicle needs to react after starting a preceding vehicle. The

Acceleration Loss Time is the time it takes for the vehicle to be brought to its final speed from a standstill.

In a further embodiment, it is provided that a current phase is terminated when a total loss time estimated for a complete phase revolution with immediate termination of the current phase is less than one estimated at any other possible remaining release time for the complete phase revolution

Total loss of time. This allows a flexible response to a changed state at the node and dynamic control of the traffic signal. If, for example, the number of vehicles changes in a currently not released inflow to the node, then the estimated total loss time can change depending on the considered time horizon. For example, was originally one

Remaining release time of 10 seconds provided and would have a state at

Not changed node, the remaining release time would be reduced to 9 seconds in the next second, etc. However, if more vehicles have been detected in a non-released inflow, this may result in a change in the estimated future loss time, since the newly added Vehicles with their lost time contribute to the total lost time. It may thus occur the situation that the originally estimated residual release duration of 10 seconds after another 5 seconds is not reduced to the remaining 5 seconds, but only to 1 second and after this second on

0 seconds, so that the current phase is canceled. This has the advantage that a current state is always taken into account at the node.

The vehicle positions are detected within the detection radius by means of suitable detectors. In this case, all known methods can be used. In particular, the traffic signal system may include suitable detectors for this purpose. However, it can also be provided that the vehicles transmit their respective vehicle position to the traffic light system within the detection radius. For this purpose, the

Traffic light system corresponding communication devices include, which receives the transmitted vehicle positions and supplies them to the controller.

In one embodiment, it is provided that the vehicle positions within the detection radius are detected and / or determined by means of floating car data and / or vehicle-to-X communication and / or camera data. This includes the

Traffic signal corresponding means for detecting and / or receiving the corresponding data.

In one embodiment it is provided that individual vehicles are described for the sake of simplicity by means of a traffic model in which the vehicles can only assume the states "waiting in the queue" or "driving at maximum speed". This simplifies the estimation of the respective

Loss times and a need for computing power required in the controller can be reduced as a result.

Parts of the device may be formed individually or in a group as a combination of hardware and software, for example as program code executed on a microcontroller or microprocessor.

The invention is based on preferred embodiments below

Referring to the figures explained in more detail. Hereby show:

1 shows a schematic representation of an embodiment of the device for the dynamic control of a traffic signal system; 2 shows a schematic representation of a time sequence of a prognosis of a phase revolution with three phases PO, P1, P2 for clarification of the method;

3 shows a schematic sequence for an arbitrary number of phases P1, P2, P, of a phase revolution to clarify the method;

Fig. 4 results for the average velocity of one by means of the method

by guided simulation in comparison to the results of conventional methods.

In Fig. 1 is a schematic representation of an embodiment of the device 1 for dynamically controlling a traffic signal system 2 is shown at a node. The device 1 comprises a controller 3 and a prediction device 4.

The controller 3 controls phases of the traffic signal system based on a loss time of the vehicles at the node. The loss time is in this case a total loss time 21 of all vehicles located in a detection radius. For this purpose, the device 1 current vehicle positions 10-x within a detection radius to the node or the traffic signal 2 are supplied. The detection of the vehicle positions 10-x can take place, for example, by means of floating-car data and / or vehicle-to-X communication and / or camera data. The device 1 may include suitable interfaces 5 for this purpose.

Based on the current vehicle positions 10-x, the prediction device 4 estimates current and future loss times of all vehicles located in the detection radius and supplies a current and estimated future total loss time 21 derived from these loss times to the controller 3. Based on the estimated current and future total loss time 21, the controller 3 controls the phases of the traffic signal system 2. The controller 3 can, for example, a

Remaining release duration 22 of the current phase of the traffic signal system 2 control.

In one embodiment, the dynamic control is based on three models. These models are a traffic model, a model of traffic signal 2 and a Loss of time model. For example, these models may be partially or completely implemented in the forecasting device 4 and / or the controller 3.

The traffic model models a behavior of the vehicles and different states of a vehicle. It uses a microscopic traffic model that looks at each vehicle individually. Each vehicle can assume only one of two states: "driving at maximum speed" (d), with the maximum

Speed of the maximum speed, or "waiting in queue" (w). This implies that more states, in particular

Vehicle tracking, lane change and braking and acceleration operations are not covered by this model. It will only be one

Considered maximum line speed, as these at nodes with

Traffic lights is limited to 70 km / h and should therefore be accessible to most vehicles. With this assumption, only the vehicle positions need to be known, not the (real) vehicle speeds.

Queues can only be established in front of a stop line of a traffic signal system. There is also the origin of a respective coordinate system of the tributaries. When a vehicle reaches the end of a queue, it is added to the queue. Queues are resolved at the beginning of the share belonging to the queue. The length L _{q of} a queue results from a number of vehicles in the queue N _q , the vehicle length l _ve h and the size of the gap l _gap between the vehicles (as length specification):

According to the formula for linear motion, the predicted distance S _ve h, t of a vehicle (state "d") to the stop line after a time t at

Maximum route speed V _ma x and an initial route S _ve h, o as follows:

'veh, t' veh, 0 max t (2)

An end of a queue is considered reached when a vehicle position S _ve h, _{r is} closer to the stop line in the model at the prediction time t than the queue end U: Syeh, t Lq (3)

When a vehicle reaches the end of the queue, the number of vehicles in the queue and thus the length of the queue is updated. In addition, the vehicle goes into the state "w". The number of vehicles in a queue N _q and the queue length l _q have no further for the traffic model

Relevance. However, they play in the loss time forecast and in the model of

Traffic signal system plays a decisive role.

The model for the traffic signal system is described below. Decisive for the resolution of a queue N _q is the release start t, the phase i. Therefore, using the model of the traffic signal system, switching times of an entire phase revolution should be estimated. Since it is an expansion criterion, it is always assumed that there is a current phase to be assessed. In this case, for a release start ti of a subsequent phase of a phase 0 to be designed, the following applies:

In this case, t _rg describes the remaining release time of the current phase, which is referred to as the remaining release time, and t o _[ _i,] a duration of a phase transition from the current phase 0 to the follow-up phase 1

Phase transitions here no release times. On the assumption that t _{rg is} known, a beginning of the subsequent phase of the phase to be measured can be determined very precisely.

However, in order to estimate the start of release of any subsequent phase i + 1 with (i> 1), the duration t _{g of} this release is necessary in addition to the start of the release of phase i. This is not known, but is estimated in this model:

T _{g, veh} corresponds to the average release time per vehicle, ie

for example, T _g , _ve h = 2 s (ie in each two seconds release time, a vehicle is degraded in a queue). It is limited downwards by the minimum release time T _g , min and up by the maximum release time T _g , _m ax. N _q corresponds to the queue of the corresponding phase i. The background to this consideration is that in most cases it makes sense to at least reduce the existing queue. However, the release time can be extended during the optimization of this phase, so that the release time with the given equation is estimated down to a certain extent. This yields the estimated start of release of a phase i + 1: + l - ti Ί ^" tg, i T ^ 0 [ί: (ί + 1)] (®)

Correspondingly, a renewed beginning of the current phase to be designed results in a fixed phase revolution, which includes i _max + 1 phases, to:

FIG. 2 shows a schematic representation of a chronological sequence of a prognosis of a phase revolution with three phases PO, P1, P2 in order to illustrate the invention. In this case, non-hatched areas indicate a release (= green traffic light) and hatched areas that are not released (= red traffic light). The individual phases PO, P1, P2 are interconnected by means of the transition times tu. The transition periods take account of safety requirements, in particular that it must be ensured that pedestrians and vehicles have sufficient time to clear the lane or the junction after a change of phase.

The loss-time model is described below. The loss time model is based on the states of the traffic model described above. Vehicles that travel at maximum speed (state "d") do not accumulate a loss time, ie in state "d" no loss time is incurred. In contrast, for every second that a vehicle lingers in state "w", one second of loss time is incurred. This results in the waiting time loss of a vehicle tv.ve _h , which has a distance S _veh, o _{, i} to the _{stop line} at which a queue with length L _q and a remaining time t, until

Release start of phase i to:

Here is the term

the remaining time of the vehicle until the

Max

Queue end, this is loss time free. The rest of the time until the

Release start t, as well as the time N _q -t _r until the queue is resolved to this position (the traffic model does not map this resolution), the vehicle waits in the queue and accumulates loss time (see Fig. 3). t _r indicates the reaction time of the vehicles to the respective predecessor in the queue. Here, it is always assumed first that a vehicle is in state "d" at the time of detection. If it is already in a queue at the given time, this is expressed by a remaining travel time of 0 s and an immediate transition to the state "w". However, this formula would give negative results and thus negative loss times when vehicles queue up to

Start of release and depletion of the queue does not reach at all (the remaining travel time would then be greater than the time until the release start and removal of the queue). However, this is not permitted, which is why 0 s is selected as the minimum loss time.

In addition, loss times are incurred at the queue resolution. In this case, the vehicles have to accelerate in reality again, even if the applied here

Traffic model does not reflect this. The resulting loss time is derived as follows: The loss time tv _{, a} or a time difference between a drive with maximum speed and an acceleration from standstill in this

Speed is given by the following equation:

s _a corresponds to the distance traveled during the acceleration process from standstill to v _max :

t _a indicates the time until v _{max is} reached: vmax

ta (1 1) as _Vjnax from equation (9) again corresponds to the distance traveled at maximum speed during the time of the acceleration process:

Overall, this results in the loss time in the acceleration from 0 m / s to v _max to:

1 ^v max

t V, a ~ 2 a (13)

Thus falls for a queue with N _q vehicles for the

Acceleration processes total following loss time tv _{, a, i} :

Further incurred loss times, in particular those caused by delays in queuing, are neglected in the context of the loss-time model and therefore not modeled.

The total loss time for a given residual release time of a current phase then results as follows. The process is shown schematically in FIG. 3 for the phases P1, P2, P.

For a given residual _{release time} t _{rg of} the current phase to be designed, a total loss time of all vehicles in the detection _radius can be estimated using the models described in the previous section. A schematic sequence of the method is shown in FIG. Here, each of the diagonal lines shown for the individual phases corresponds to a trajectory of a vehicle position 10-1,..., 10-7 with respect to a vertical time axis and a horizontal position axis. The procedure now proceeds as follows: First, the start of release ti of the phase following the current phase is estimated according to equation (4). On the basis of the estimated start of release, it is possible to use the iteration above all in the

Detection radius located vehicles (step 101) of the considered phase with equations (1), (2) and (3) the number of vehicles in the queue and the queue length are estimated. Here, in each of the iteration steps for the respective vehicle according to equation (8) a resulting loss time tv. _veh is calculated (step 102). Fall for the full queue

then loss times according to equation (14) (method step 103). In addition, the duration of the phase t _{g is} estimated by equation (5) (step 104).

With the estimated release time, by means of equation (6) again the

Approval start of the next phase are estimated (step 100), concomitantly again the queue length, the loss times and the release period. The process continues until a phase revolution is completed and estimated by equation (7), the re-release start of the current phase. Here a distinction is made: If a vehicle can still pass the _{stop line} during the originally remaining release time t _rg , it does not accumulate a loss time. Otherwise, the loss time is calculated according to equation (8) (method step 102). Again, then fall in addition to the above

Acceleration loss times for the resolution of the queue according to equation (14).

The estimated total loss time tv (t _rg ) as a function of the remaining release time t _rg then corresponds to the sum of all loss _times calculated during the process described here.

Based on this estimated total loss time, the controller controls the timing of each phase of the traffic signal.

In the previous section it was described how the total loss time tv (t _rg ) at the node can be estimated for a given residual _{release time} t _rg . A single value for the total loss time is still no indication of how well a

Switching point is. The control approach consists in comparing the total loss times resulting from the different residual release times and so on to determine an optimal residual release time. Since this is an expansion criterion, it is only interesting if the current phase should be aborted or not.

The current phase should be aborted if the following inequality is met:

This inequality expresses that the controller aborts the current phase when canceling and switching to the subsequent phase in the current second will cause less total lost time than any switch to any other possible one

Rest release time. As soon as even a single estimated future total loss time is found with t _rg > 0 s, which causes less total loss time than t = 0 s, the current phase is extended with a residual release time> 0 s and the procedure is repeated in the next second. The maximum residual _{release time} t _rg , _m ax is calculated here from the previous release time t _{g of} the current phase and the maximum release time

Tg, max · trg nax ^ g.max tg (6)

This ensures that the release time can not be extended beyond the maximum release time. In addition, the described optimization is only started when the minimum release time T _g , _m m has been reached, so that the lower limit is maintained in each case.

Overall, the control receives the estimation of the forecasting device the

Ability to estimate the impact of tax decisions on traffic at the hub. The advantage of this method and the device is that for this purpose no communication with adjacent traffic signal systems or a central processing unit is necessary.

The biggest advantage of the device and the method is that the loss times of the vehicles are used as a direct decision variable. As a result, the loss time is not only used for quality assessment, but also directly for control.

Furthermore, the method and the device have the advantage that they have an often applied time-dependent, but fixed phase influence are designed traffic-dependent. This can be reacted directly to changing traffic conditions at the hub, which favors the flow of traffic.

This is reflected in shorter travel times with less lost time

Road users, as will be described in the following section and with reference to FIG. 4.

The control is based on the current traffic situation and leads to a more effective release time allocation in the context of a phase revolution, since the weighting of a phase is directly related to the traffic volume. Phases with large traffic streams accumulate faster loss times and are considered by means of the described method and the device described rather than weak demand traffic streams, which then receive correspondingly low release times.

Despite the simple design of the method and the device, these are real-time capable, since no computation-intensive methods are necessary, as is partially the case with other model-based methods.

Since the described solution is a decentralized independent process, no additional communication infrastructure needs to be set up. This saves effort and costs.

The method described and the device described relate

Interactions at network level with in the local control of the traffic light system, without having to resort to static approaches such as master plans, etc. The dynamic control of traffic lights is thereby significantly improved.

FIG. 4 shows results for the mean velocity of a simulation carried out by means of the method in comparison to the results of conventional methods.

As part of the simulation, different network sections were controlled using the described method (VZP). For comparison, was also a waiting time optimal Fixed time control (FZS), one based on the fixed-time control

Framework control (RPS) and a classical time gap control (ZLS) considered. The modeling and the simulation have the following properties:

The route includes five consecutive nodes. The access roads or tributaries are each 500 m long.

The wait-time-optimal fixed-time control is calculated, for example, according to the manual for the design of road traffic facilities (Research Association for Roads and Transportation, FGSV Verlag, Cologne, 2015) as a standard method.

The nodes are evenly distributed once and unevenly. Accordingly, the fixed time and frame plan control coordinates once in both and once in only one direction. The orbital period is prescribed for these two procedures with 60 s; in the regular distribution, the nodes are at the sub-point distance.

The scenarios are further subdivided into scenarios with and without bending currents. For turning left turn are in the appropriate

Scenarios a left turn lane and a phase to the secured

Left turn introduced.

The networks are burdened with the utilization of 60%, 85% and 100%.

The phase sequence is fixed and unchanging. It is the same for all tested approaches.

As test vehicles only passenger cars are used.

The minimum permissible release time is 5 s, the maximum release time 90 s.

The constant, permissible maximum speed is 50 km / h.

The critical time gap for the time gap control is defined as 2 s. It is detected at detectors 20 m before the stop line of the corresponding access road.

As planning traffic strengths, 80% of mainstream traffic is examined to 20% in secondary traffic (80-20). In addition, the

Consider ratios 90-10 and 70-30 without adjusting the fixed-time control or the master plan. This is supposed to be the deviation of the actual

Traffic intensity of the planning traffic strength simulate and check the flexibility of the procedures. The simulation time is 1 h for each simulation run, whereby the measured values of the first hour are discarded because the first hour should only be used for grid filling.

For the evaluation, the average speed in the entire network is considered. It results from the sum of all driven distances in the net and the

Total duration of all rides.

Figure 4 shows the mean velocities in the network in the different scenarios compared to the conventional approaches. The y-axis corresponds to the mean velocities in the network, the x-axis indicates the respective scenario. Shown are the individual quartiles, where the upper quartile is so small that it is in this

Representation is not visible.

In Fig. 4, N1 denotes the network with regular node spacings without intersecting currents, N2 the network with irregular inter-node distances without intersecting currents, and N3 and N4 are each those with intersecting currents. From the results, the following conclusions can be drawn:

The fixed-time control (FZS) works particularly well with less complex nodes with low to medium utilization. At heavy load, traffic often collapses and in some cases only medium

Speeds of less than 10 km / h.

The master plan control (RPS) is so flexible that it can absorb fluctuations relatively well. By coordinating the core release times on the coordinated line, the overall result is even higher

Speed achieved with the pure time gap control (ZLS).

The time gap control works basically according to the principle of

Backlogs, which is why the coordination with free-runners barely works. However, the sidestreams are considered more closely, which is why even the time-gap control without a framework plan in the grid analysis delivers partly better results than the fixed-time control. In addition, the controller is flexible enough that there are no overloads.

The control unit based on the described method (VZP)

To a certain extent, the advantages of the time gaps and master plan control: It is not bound to fixed round-trip times, but still brings with it the network problem into the controller. In this way, in all scenarios better results can be achieved by the controller based on total time lost than with the other methods. The average speed in the network is here increased on average by 5% to 25% compared to the usual methods (see Table 1).

Table 1: Relative change in mean speed when using the

described method or the described device (VZP).

The described method and the described device can be used in the field of control systems of traffic signal systems. This includes virtually all public highway technology that is used to regulate traffic flows. In particular, the field of traffic-dependent traffic signal control here increasingly importance is attached, since there are significant potential for saving loss times, fuel and pollutant emissions by the road users. The integration of new characteristics, e.g.

Loss times, is of particular interest for light signal manufacturers and municipalities, in order to be able to process the increasing volume of traffic in the future in an appropriate manner.

LIST OF REFERENCE NUMBERS

1 device

2 traffic signal system

3 control

4 forecasting device

5 interface

10-x vehicle position

21 total loss time

22 remaining release time

PO phase

P1 phase

P2 phase

Pi phase

100-104 process steps

Claims

claims

1. A method for dynamically controlling a traffic signal system (2), wherein

Phases (PO, P1, P2, Pi) of the traffic signal system (2) are controlled by means of a controller (3) on the basis of a loss time,

characterized in that

the loss time is a total loss time (21) of all vehicles in a detection radius, the total loss time (21) being estimated by means of a forecasting device (4) based on current and estimated future vehicle positions (10-x) of the vehicles in the detection radius.

A method according to claim 1, characterized in that for estimating the total loss time (21) by means of the forecasting means (4), a number of vehicles in a queue and a queue length are estimated based on the individual vehicle positions (10-x).

3. The method according to claim 2, characterized in that for estimating the total loss time (21) by means of the forecasting device (4) a

Remaining release duration (22) of a phase (PO, P1, P2, Pi) and a release start of a phase following this phase (PO, P1, P2, Pi) (PO, P1, P2, Pi) based on the number of vehicles in the Queue is estimated.

4. The method according to any one of claims 3, characterized in that for

Estimating the total loss time (21) by means of the forecasting device (4), a loss time for each vehicle in the detection radius is estimated on the basis of the remaining release duration (22), wherein for the loss time of one

Vehicle is considered in each case a waiting time loss, a reaction loss time and an acceleration loss time.

5. Method according to one of claims 1 to 4, characterized in that a current phase (PO, P1, P2, Pi) is ended when a total loss time (21), estimated for a complete phase revolution, upon immediate termination of the current phase (PO, P1, P2, Pi) is less than any other possible one Remaining Release Time (22) Estimated Total Loss Time (21) for the entire phase revolution.

6. The method according to any one of the preceding claims, characterized in that the vehicle positions (10-x) are determined within the detection radius by means of floating car data and / or vehicle-to-X communication and / or camera data.

7. The method according to any one of the preceding claims, characterized in that individual vehicles are described for simplicity by means of a traffic model in which the vehicles can only assume the states "waiting in the queue" or "driving at maximum speed".

8. A device (1) for dynamically controlling a traffic signal system (2), comprising: a controller (3), wherein the controller (3) is designed such, phases (PO, P1, P2, Pi) of the traffic signal system based on a loss time steer, characterized by

a forecasting device (4),

the loss time being a total loss time (21) of all in one

Detecting vehicle, and wherein the prediction means (4) is adapted to estimate the total loss time (21) based on current and estimated future vehicle positions (10-x) in the detection radius.