Method and device for traffic flow control

Abstract

According to a first aspect of the invention, a method for traffic flow control is proposed. The method comprises generating and transmitting driving behavior instructions for at least one road user, in particular a connected vehicle. The driving behavior instructions are generated by creating a digital model of a traffic corridor using current traffic conditions, simulating a traffic flow based on this digital model, and determining the driving behavior instructions based on this traffic flow simulation.

Description

[0001] The invention relates to a method and a device for traffic flow control. The device can, in particular, be an infrastructure system for driving support of at least partially automated, networked motor vehicles. The invention further relates to a computer program. State of the art

[0002] As in other technology areas, connectivity is playing an increasingly important role in vehicle applications. More and more vehicles are able to connect with other road users, infrastructure components (e.g., so-called roadside units), or backend services in a cloud. This is where so-called V2X or V2V communication comes into play.

[0003] Radio communication standards for these purposes include Wi-Fi-based direct communication standards (DSRC, IST-G5) or approaches based on mobile communication, both in a direct and indirect form (C-V2X). In particular, the networking of vehicles with infrastructure systems is gaining importance. Such systems can support vehicles in their driving tasks by, for example, using roadside sensors to provide environmental information that the vehicle's on-board sensors cannot generate themselves, or can only generate to a limited extent. This support can even extend to direct intervention in driving functions, for example, when partially or fully automated vehicles are no longer able to continue driving or are only able to do so to a limited extent due to a limited view of the surroundings.This networking of vehicles and infrastructure systems via V2X communication can significantly improve both road safety and traffic efficiency.

[0004] Another possibility for influencing the maneuvers of partially or fully automated vehicles from outside the vehicle is for several such vehicles to network directly with each other and exchange their environmental information and driving tasks, thus enabling cooperative behavior between the vehicles. This offers advantages particularly in scenarios where the connected vehicles interact directly with each other (e.g., lane changes on the highway) and less so at large, confusing traffic junctions, where even the combined field of vision of several vehicles is not guaranteed to be sufficient to oversee the entire traffic situation.

[0005] German Patent Application DE 10 2020 102 329 A1 discloses a method for controlling a vehicle in a traffic situation in which a vehicle merges into a traffic flow, minimizing the overall energy consumption. The method records traffic data from the vehicles involved in the traffic situation, and generates at least one control signal for at least one of these vehicles to assist in merging. This can be achieved using, for example, traffic data and / or sensor data from environmental sensors of the vehicles involved and / or from a vehicle-external traffic monitoring device, which may be provided as part of a local infrastructure.

[0006] A traffic flow advisory system is known from published patent application DE 10 2010 050 805 A1. The traffic flow rate of several lanes is calculated and communicated to a vehicle driver. The traffic flow rate can be enhanced with data transmitted wirelessly from infrastructure sensors. In one embodiment, the system, as an interface between driver and machine, can advise the driver based on predetermined criteria as to which lane is most suitable for their driving behavior. Such advice can include directional visual, acoustic, or haptic cues.

[0007] When a certain traffic intensity exceeds a road's capacity, traffic becomes denser and begins to slow down. Under such conditions, the phenomenon of traffic waves, also known as braking shock waves or "stop and go," can occur. A related problem is detecting the "end of a jam," where road users must suddenly brake sharply from a normal speed.

[0008] Current measures to counteract these phenomena include variable speed limits. This requires complex structures and can only be implemented with a limited temporal and spatial frequency.

[0009] An existing solution for traffic jam warnings involves exchanging driving data between users of navigation software on smartphones. This process aggregates data from multiple journeys and uses it to detect traffic jams. However, this aggregation method limits the reactivity and precision of this solution and is insufficient for better handling of braking shock waves and end-of-traffic situations. Disclosure of the invention

[0010] It can therefore be considered an object of the invention to provide a reliable method for traffic flow control in which phenomena such as braking shock waves or suddenly occurring traffic jams are effectively prevented.

[0011] It can be considered a further object of the invention to provide a device for traffic flow control, in particular an infrastructure system for driving support of at least partially automated networked motor vehicles, which has a high level of reliability and by means of which phenomena such as braking shock waves or suddenly occurring traffic jams are effectively prevented.

[0012] According to a first aspect of the invention, a method for traffic flow control is proposed. The method comprises generating driving behavior instructions for at least one road user, in particular a networked motor vehicle, and sending them to the at least one road user, in particular to the networked motor vehicle. The driving behavior instructions are generated in such a way that a digital model of a traffic corridor is generated using current traffic conditions, a traffic flow is simulated based on the digital model, and the driving behavior instructions are determined based on the traffic flow simulation.

[0013] According to a second aspect of the invention, a traffic control unit is proposed which is designed to generate driving behavior instructions for at least one road user, in particular a networked motor vehicle, wherein the driving behavior instructions are generated in such a way that a digital model of a traffic corridor is generated by means of current traffic conditions and a traffic flow is simulated based on the digital model and driving behavior instructions are determined based on the simulation of the traffic flow.

[0014] According to a third aspect of the invention, a device for traffic flow control is proposed, which is designed in particular as an infrastructure system for driving support of at least partially automated networked motor vehicles, and which is designed in particular to carry out a method according to the first aspect. The device comprises at least one communication unit, at least one traffic control unit according to the second aspect, and at least one road traffic sensor, wherein the current traffic conditions are detected in particular by means of the road traffic sensor.

[0015] The invention is therefore based on cooperative maneuver planning of the vehicles involved. This not only makes it easier for individual vehicles to deal with such a traffic situation, but also prevents the occurrence of such braking shock waves and traffic jam situations, to the benefit of all road users.

[0016] For the same reason, this approach offers advantages beyond a simple V2X traffic jam warning, as these cannot be received by all road users, nor do they significantly reduce the phenomena mentioned.

[0017] Especially in the case of braking shock waves in slow-moving traffic, this is a complex, nonlinear phenomenon in which the traffic flow as a whole can be viewed as a single, oscillating system. Centralized traffic control is essential to detect the onset of oscillations and to prevent them through a unified strategy.

[0018] According to the invention, collective driving behavior advice is used. Unlike vehicle-specific trajectory recommendations, the utilization of the communication channel remains stable, regardless of the number of road users involved. This type of communication places lower hardware and software requirements on the participating vehicles and infrastructure systems, both for communication (multicast) and for the driver assistance systems.

[0019] The driving behavior instructions according to the invention can also be implemented by non-partially or fully automated motor vehicles, thus the method described here is also suitable for mixed traffic conditions, in addition to the existing traffic control options. Furthermore, higher-level planning enables the planning algorithm to be specialized to the specific conditions of the traffic axis on which it is applied. The generation, transmission, and use of the driving behavior instructions can, in particular, be defined in such a way that an infrastructure operator is able to influence the movement of several networked motor vehicles, but does not have to assume responsibility for the movement of the vehicles themselves.This is a decisive advantage of the invention over the prior art, since reliable control of a vehicle from the infrastructure, and thus the assumption of responsibility for vehicle movement by the infrastructure, is technically very complex and organizationally and legally difficult to implement. The latter is due to the fact that different companies will usually develop and sell the infrastructure system and the AD vehicle, which cannot and do not want to mutually assume responsibility for the correct behavior of each other's product.

[0020] The term "connected motor vehicle" encompasses a motor vehicle that has a suitable communication device with which the connected motor vehicle can exchange data with other road users, in particular with an infrastructure system. For this purpose, a wireless data connection is established, via which the connected motor vehicle can send and / or receive data. This can preferably be a radio connection, for example, a mobile radio connection or a direct wireless connection. Such communication between a motor vehicle and another road user is also referred to as V2X or C2X communication.

[0021] The term “at least partially automated” covers one or more of the following cases: assisted driving, partially automated driving, highly automated driving, fully automated driving of a motor vehicle.

[0022] Assisted driving means that a driver of a motor vehicle continuously performs either the lateral or longitudinal control of the vehicle. The other driving task (i.e., controlling the longitudinal or lateral control of the vehicle) is performed automatically. This means that with assisted driving, either the lateral or longitudinal control is controlled automatically.

[0023] Partially automated guidance means that the longitudinal and lateral guidance of the vehicle is automatically controlled in a specific situation (for example, driving on a highway, driving within a parking lot, overtaking an object, driving within a lane defined by lane markings) and / or for a specific period of time. The driver of the vehicle does not have to manually control the longitudinal and lateral guidance of the vehicle. However, the driver must continuously monitor the automatic control of the longitudinal and lateral guidance in order to intervene manually if necessary. The driver must be ready to fully assume control of the vehicle at any time.

[0024] Highly automated guidance means that for a certain period of time in a specific situation (for example: driving on a highway, driving within a parking lot, overtaking an object, driving within a lane defined by lane markings), the longitudinal and lateral guidance of the motor vehicle are automatically controlled. The driver of the motor vehicle does not have to manually control the longitudinal and lateral guidance of the vehicle. The driver does not have to constantly monitor the automatic control of the longitudinal and lateral guidance in order to be able to intervene manually if necessary. If necessary, a takeover request is automatically issued to the driver to take over control of the longitudinal and lateral guidance, in particular with a sufficient time reserve. The driver must therefore potentially be able to take over control of the longitudinal and lateral guidance.Limits of the automatic control of lateral and longitudinal guidance are automatically detected. With highly automated guidance, it is not possible to automatically achieve a minimal-risk state in every initial situation.

[0025] Fully automated guidance means that in a specific situation (e.g. driving on a motorway, driving within a parking lot, overtaking an object, driving within a lane defined by lane markings), the longitudinal and lateral guidance of the vehicle are automatically controlled. The driver of the vehicle does not have to manually control the longitudinal and lateral guidance of the vehicle. The driver does not have to monitor the automatic control of the longitudinal and lateral guidance in order to intervene manually if necessary. Before the automatic control of the lateral and longitudinal guidance is terminated, the driver is automatically prompted to take over the driving task (controlling the lateral and longitudinal guidance of the vehicle), particularly with a sufficient time reserve. If the driver does not take over the driving task, the vehicle is automatically returned to a state with minimal risk.Limits of the automatic control of lateral and longitudinal guidance are automatically detected. In all situations, it is possible to automatically return to a system state with minimal risk.

[0026] Driverless control or guidance means that, regardless of a specific application (for example, driving on a highway, driving within a parking lot, overtaking an object, driving within a lane defined by lane markings), the longitudinal and lateral guidance of the vehicle are automatically controlled. The driver of the vehicle does not have to manually control the longitudinal and lateral guidance of the vehicle. The driver does not have to monitor the automatic control of the longitudinal and lateral guidance in order to intervene manually if necessary. The vehicle's longitudinal and lateral guidance are thus automatically controlled, for example, on all road types, speed ranges, and environmental conditions. The driver's entire driving task is thus automatically taken over. The driver is therefore no longer required.The vehicle can therefore drive from any starting position to any destination without a driver. Potential problems are resolved automatically, without driver assistance.

[0027] Remote control of a motor vehicle means that the lateral and longitudinal guidance of the vehicle is remotely controlled. This means, for example, that remote control signals are sent to the vehicle to remotely control the lateral and longitudinal guidance. Remote control is carried out, for example, using a remote control device.

[0028] For the digital model of the traffic corridor, data from infrastructure-based sensors can be collected, representing current traffic conditions and processed together with topological information about the traffic corridor. For example, the data can originate from V2X "Cooperative Perception Messages" (CPM) from connected road users. Data from infrastructure-based sensors, e.g., traffic monitoring cameras, radar, lidar, and induction loops, is another data acquisition option. The data can map the observed traffic over time as so-called "object tracks."

[0029] The topological information of the traffic corridor is preferably contained in a digital map of a road network. The digital map can, for example, be entered manually or obtained from a map provider. Alternatively or additionally, the topological information of the traffic corridor can be determined from perception data received from a plurality of networked motor vehicles, in particular using machine learning. For this purpose, for example, frequent traffic patterns and trajectories of passing vehicles can be aggregated and automatically converted into a topological road model.

[0030] Traffic flow simulation is preferably based on a real-time, predictive model with rule-based traffic control. Such a model can, for example, incorporate a machine learning (ML) approach, such as a convolutional neural network (CNN) or a recurrent neural network (RNN). The advantage of this predictive model is that it is capable of accurately predicting "stop-and-go braking shock waves" and the ends of traffic jams, allowing for counteracting these.

[0031] The simulation preferably implements a machine learning approach, with the machine learning model trained using historically recorded traffic conditions. The entire model can consist of a CNN, RNN, or similar. An example of a function that can be trained in this way is the estimation of the probability of a braking shock wave occurring within the next minute based on the current traffic situation.

[0032] Alternatively, the model can be based on physical modeling, e.g., numerical simulation. For example, physical modeling involves dividing the road topology into cells, which can then serve as the basis for numerically solving differential equations in a traffic flow model. Variables such as the number of vehicles or average speed in each cell are simulated over time. Another physical modeling option is agent-based traffic simulation. This model models the behavior and interactions of all individual road users (microscopic simulation). In the case of physical modeling, specific model parameters can be adjusted, for example, using a multivariate optimization procedure.An example of an optimizable variable is the average reaction time of a human driver when faced with braking traffic ahead.

[0033] A combination of physical modeling and ML approaches can also be used, a so-called "hybrid model." Here, certain relationships are learned within the model and then used in a physical simulation model. Relevant functions that can be learned include, for example, terms of the transport equations that ensure the spatial and temporal propagation of the simulated variables (e.g., traffic density and average speed per cell) in a macroscopic traffic simulation. These are typically nonlinear and depend on many factors that may only be partially captured by traditional models, such as the reaction time of a human driver when faced with braking traffic ahead under real-world conditions.These terms can be designed entirely as functions supported by a neural network (also known as RNNs or CNNs) to learn the influence of a multitude of (possibly hidden) parameters. Inputs include, for example, the traffic situation along the traffic corridor and other relevant conditions, such as the weather or the time of day.

[0034] For all these types of models, the use of real traffic data to adjust and adapt the model can bring advantages.

[0035] In a preferred embodiment of the invention, the real-time predictive model is used to determine optimal driving behavior for specific sections of the traffic corridor, and driving behavior advice is derived from this. The driving behavior advice includes, in particular, speed recommendations and / or acceleration and deceleration targets, and / or recommendations for maintaining distances from the vehicle ahead and / or recommendations for using a specific lane.

[0036] To prevent the occurrence of traffic jams and braking shock waves, simple rules can be applied after an imminent occurrence of an event has been identified: The following rules may apply to driving behavior instructions, for example: - The deceleration must be anticipated instead of trying to keep the distance to the car in front constant. - Acceleration should remain moderate - The target speed must not significantly exceed the average speed to avoid repeated braking.

[0037] The extent to which these rules apply can be determined in advance using simulations or based on empirically obtained values. One challenge is finding values ​​that have a sufficiently broad range of applicability.

[0038] In a preferred embodiment of the invention, the driving behavior advice includes at least speed recommendations. It has been shown that even small adjustments in speed can effectively prevent traffic jams or braking bumps; thus, a speed recommendation, especially a maximum speed, is particularly effective for achieving optimized traffic flow. The speed recommendations can, for example, alternatively or additionally include a speed interval and / or a minimum speed.

[0039] In a preferred embodiment of the invention, the driving behavior advice is calculated and transmitted specifically for a particular lane. This has the advantage that events such as braking bumps or sudden traffic jams can also be prevented on a lane-specific basis. Furthermore, it is advantageous if the driving behavior advice includes lane change recommendations.

[0040] In a preferred embodiment of the invention, the simulation comprises a self-learning algorithm, whereby a continuous optimization of the traffic flow control takes place during operation.

[0041] An alternative to predefined rules for traffic control are self-learning control algorithms based on, for example, Reinforcement Learning (RL), Deep Q Learning (DQL), or similar. In this case, the automated traffic control is continuously optimized by the control unit. The process is divided, for example, into several phases. After a data acquisition phase, a simulation model can be initialized. The simulation is used as a training environment for, for example, the RL process, in which the model can integrate the control commands of the traffic control unit into the simulation process. The simulation thus serves to generate synthetic data for the initial training of the traffic control system. The simulation requires traffic flow data as boundary conditions, which can include both real recorded traffic data and synthetically generated scenarios.The effect of the control in the simulation can be evaluated in many scenarios.

[0042] The conditions that occur in the simulation are evaluated using a so-called "reward function." This function is defined in such a way that, for example, the traffic speed is maintained as consistently as possible, and traffic jams and other dangerous situations are mitigated.

[0043] By testing possible actions (“policies”) in different conditions, the algorithm learns which sequence of actions leads to the overall better result over time. To gather new information, the algorithm can deviate to some extent from the previously known best control. The degree of exploration tendency can typically be input into RL algorithms (“exploration-exploitation trade-off”). An initial trial phase (“bootstrapping”) is best carried out using a simulation. This way, an unfavorable control sequence, which can provide valuable information for learning, is not applied to real traffic.

[0044] After the offline training phase has been completed, and adequate (but not yet optimal) traffic control has been implemented, the system can switch to live operation. In this live operation, the control system can be further optimized through online learning. For live operation, the best known control system is followed more closely, so learning is slower, but good control is always offered. Such a self-learning process allows the control system to be adapted to real local conditions. Thus, the quality of a model only plays a role at the beginning; later, experience from online learning predominates. This is advantageous because the phenomena to be treated in reality depend on many factors that are not easy to model. For example, traffic rules are not uniform across geographies, driving styles are sometimesculturally influenced, special road topology, temporary disruption to traffic occurs.

[0045] In a preferred embodiment of the invention, the driving behavior instructions are transmitted to a plurality of road users using a multicast and / or broadcast communication method. A multicast or broadcast communication method is particularly advantageous in slow-moving traffic. This allows for lower utilization of the radio communication channel compared to a unicast communication method, since a single message can control a large number of vehicles without having to address them individually.

[0046] To ensure the message is sufficiently specific to efficiently regulate traffic along the corridor, recipients can be addressed in different ways. One option is to adjust the transmitter's transmission power so that only affected vehicles can receive the message. Another option is to specify a geographical validity zone in the message. A validity zone can be represented, for example, by a quadrilateral or a polygon, with the vertices labeled with absolute coordinates. Absolute coordinates can correspond, for example, to the UTM or WGS84 standard. In the case of a network of radio communication units, validity zones can be addressed by the nearest radio communication unit using a mesh method. The validity zones must be checked by the recipients to determine the relevance of the message.

[0047] The message specifically conveys information regarding speed control and the distance to the vehicle in front. It can also include information regarding lane change frequency, lane selection, or yielding priority in the zipper procedure. The target vehicles can then incorporate these instructions into their driver assistance systems, or alternatively, into their trajectory planning for automated vehicles. The generated driving behavior instructions can be sent to the respective vehicles via V2X communication, along with environmental information.

[0048] In a preferred embodiment of the invention, the driving behavior instructions are received by a networked motor vehicle and output to the driver of the motor vehicle, particularly audiovisually and / or haptically. This advantageously allows the invention to be implemented in mixed traffic involving manual and partially or fully automated motor vehicles.

[0049] Particularly relevant driver assistance systems that can be used to implement driving behavior advice include Automated Cruise Control (ACC) or UX measures that support a human driver in maintaining a certain speed. These include: - Warnings in visual form, e.g. in the cluster display. - Acoustic warnings - Haptic control elements such as accelerator pedal or steering wheel, ActivePedal or related products

[0050] Suitable UX devices will become increasingly common in future motor vehicles, as they are intended to implement the EU regulation on "Intelligent Speed ​​Assist." This invention proposes an additional use for these UX devices.

[0051] An important feature of the invention is that motor vehicles voluntarily follow the infrastructure instructions, provided they are functionally and safely feasible. Since the responsibility for safety lies with the motor vehicles, they must always verify the plausibility of external instructions and evaluate them against their own planning before implementing them. In the case of driver assistance systems in manually controlled vehicles, these must be designed so that they do not impair the driver's judgment. To ensure good vehicle acceptance of infrastructure instructions, they should only deviate from the usual traffic situation within reasonable limits. In particular, the maximum permitted speed must never be exceeded.Thus, the vehicle-side safety concept remains unchanged, and the infrastructure does not have to assume any additional safety requirements (in the sense of functional safety). Short description of the characters

[0052] Embodiments of the invention are described in detail with reference to the accompanying figures. Fig. 1 shows an example of a system architecture for the generation and transmission of driving behavior information from an infrastructure system to networked motor vehicles according to an embodiment of the invention. Fig. 2 shows a flowchart of a traffic control according to an embodiment of the invention. Fig. 3 shows a flowchart of a self-learning traffic control according to an alternative embodiment of the invention. Fig. 4 illustrates the passive recording of traffic data in an embodiment of the invention. Fig. 5 illustrates an automated learning of a traffic model from measured traffic data within the framework of an embodiment of the invention. Fig. 6 illustrates a bootstrapping method of a self-learning traffic control unit by means of an interactive traffic simulation (offline learning) according to an alternative embodiment of the invention. Fig. Figure 7 shows an example of a modeling of traffic jam ends and braking shock waves. Fig. 8 illustrates the operation of a self-learning traffic control system in accordance with an embodiment of the invention. Fig. Figure 9 shows an example scenario for slow-moving traffic on a motorway-like road. Preferred embodiments of the invention

[0053] In the following description of the exemplary embodiments of the invention, identical elements are designated by identical reference numerals, whereby a repeated description of these elements is omitted where appropriate. The figures only schematically illustrate the subject matter of the invention.

[0054] In Fig. Figure 1 shows a possible system architecture 100 for generating and transmitting driving behavior information from an infrastructure system to networked motor vehicles according to an exemplary embodiment of the invention. A road traffic sensor 102, for example a camera, detects the current traffic and sends data to a first roadside unit 104. The roadside unit 104 also receives CAM / CPM data from networked motor vehicles 112, 114, 116 via V2X communication and is coupled to a traffic control unit 106, which is configured to generate driving behavior information for at least one road user, in particular a networked motor vehicle.The driving behavior advice is generated by using current traffic conditions detected by the road traffic sensor 102 to generate a digital model of a traffic corridor. Based on the digital model, a traffic flow is simulated, and driving behavior advice is determined based on the traffic flow simulation. The driving behavior advice is transmitted to the roadside unit 104 and, if necessary, to other roadside units 108. The roadside units 104, 108 communicate the driving behavior advice to the networked motor vehicles 112, 114, 116 using a multicast communication method.

[0055] In Fig. Figure 2 shows a flowchart 200 of a traffic control system according to an exemplary embodiment of the invention. In a first step 202, passive data acquisition of current traffic data takes place. In step 204, a predictive model for simulating traffic flow is operated using the acquired data. From this, driving behavior advice for traffic flow control is generated in step 206.

[0056] In Fig. Figure 3 shows a flowchart 300 of a traffic control system according to an exemplary embodiment of the invention. In a first step 302, passive data acquisition of current traffic data takes place. In step 304, a self-learning model for simulating the traffic flow is started using the acquired data. Using offline acquired training data and a bootstrapping method, a first model for simulating the traffic flow is generated in step 306. From this, driving behavior instructions for traffic flow control are generated in step 308. In parallel, the model is further improved through real-time learning in step 310.

[0057] Fig. Figure 4 illustrates passive data acquisition. Data 402 from infrastructure-based sensors and data 404 collected from various connected road users, such as connected motor vehicles, are aggregated in an object tracking and fusion unit 406, and the results are made available in a database 408.

[0058] Fig. Figure 5 shows the process of automated learning of a traffic model from measured traffic data. The data is stored in a database 408 (see Fig. 4). In a first path 502, the data is initially aggregated in step 504, from which a topological map 506 of the relevant traffic corridor is generated. In parallel, in a step 508, a traffic flow simulation and synthetic traffic scenarios 510 and simulation parameters 512 are generated, e.g., using regression or ML approaches. The synthetic traffic scenarios 510, the topological map 506, and the simulation parameters 512 are used in step 514 by a predictive model or a simulation engine to obtain an improved traffic model.

[0059] Fig. 6 shows the process of bootstrapping a self-learning traffic control unit using an interactive traffic simulation (offline learning). A simulation unit 902 delivers initial, simple results offline (i.e., not during ongoing operation) and makes them available to the self-learning traffic control unit 906 via an interface 904. The traffic control unit comprises a learning module 908. The delivered results are evaluated by the learning module 908 via a reward function 910 and incorporated into a utility function 912. From these results, driving behavior instructions or a traffic control strategy ("policy") are generated in an ML module 914, e.g., using deep Q learning, and returned to the simulation unit 902 via a further interface 916.The simulation unit 902 can thus calculate an improved simulation of the traffic flow and better results, which in turn can be passed on to the self-learning traffic control unit 906 for the purpose of further refining the simulation.

[0060] Fig. Figure 7 illustrates a macroscopic traffic simulation model for traffic jam ends and braking shock waves. For the model, the road topology 610 of the traffic corridor under consideration is divided into individual, discrete cells 601, which, for example, have a certain, fixed size. For each cell 601, it can be calculated how certain traffic flow parameters, e.g., traffic density, the number of vehicles per cell, the average speed, etc., change over time. Fig. Figure 7 shows the road topology 610 at a first time 612 and a second, later time 614. In this example, a dark color of the cell represents a low average speed. In the example shown, it is evident how a zone 622 (a contiguous area of ​​cells) with a comparatively low average speed expands against the direction of travel 680, so that at the later time 614, further zones 626, 628 with a comparatively low average speed have emerged. This is a so-called braking shock wave or braking wave. The macroscopic traffic simulation model can predict such effects by numerically simulating the cells 601.

[0061] Fig. 8 shows the processes of a self-learning traffic control system within the scope of an exemplary embodiment of the invention, including an "online learning" method. Data is collected using an infrastructure-side sensor system 702, wherein the data represents current traffic conditions. Furthermore, V2X data is received from a plurality of road users 704, which includes topological information of the traffic corridor. An environmental model is created in a fusion module 708 from the data from the infrastructure-side sensor system 702 and the topological information. The environmental model is made available to a self-learning traffic control unit 706 via an interface 714. The self-learning traffic control unit 706 comprises a learning module ("learning engine") 703 in which various ML methods can be implemented.Based on an initially learned utility function 724, the learning module 703 proposes one or more traffic control actions in the form of a traffic control strategy (“policy”) 726. Driving behavior instructions for the road users 704 are derived from the traffic control strategy 726 and transmitted to them via an interface 712 using V2X communication. The road users 704 and the infrastructure-side sensors 702 now collect data again, from which a new environmental model is created by the fusion module 708 and made available to the self-learning traffic control unit 706. The learning module 703 can now determine, using a reward function 722, whether the utility function 724 and / or other parameters need to be adjusted in order to improve the policy 726. A self-learning traffic control system is implemented via the feedback loop thus established.

[0062] Fig.9 shows, by way of example, a traffic corridor 800 in which traffic flow is controlled according to a possible embodiment of the invention. The traffic corridor 800 comprises a three-lane roadway 810, with road users 823, 834 moving along the direction of travel 818 on the three lanes. The traffic corridor 810 is divided into several sections 822, 824, 826, 828. At the time shown, traffic is flowing in cells 822, 824, 826, while traffic is already backed up in cell 828. According to the invention, the formation of braking shock waves and the resulting "stop-and-go" traffic should be prevented or at least mitigated. For this purpose, the traffic flow can be simulated and optimized for each of the sections 822, 824, 826, 828 by dynamically generating specific driving behavior instructions for each section 822, 824, 826, 828.The driving behavior advice can include, for example, recommendations for speeds, distances, and / or lane guidance. To record current traffic conditions within the cells, such as traffic density and / or average speed, several infrastructure units 804, 808 are arranged stationary along the traffic corridor 810, each having a communication unit. Furthermore, at least one of the infrastructure units 804, 808 has at least one road traffic sensor. Using their respective communication unit, the infrastructure units 804, 808 can send data via V2X communication to the road users 832, 834 and / or receive data from the road users 832, 834. The sent data can be, for example, driving behavior advice 816 and / or CPM ("Cooperative Perception Messages") 814.Using their respective communication units, the infrastructure units 804, 808 can further exchange data with a central traffic control unit 806. The traffic control unit 806 is configured to generate respective driving behavior instructions for the sections 822, 824, 826, 828, wherein the driving behavior instructions are generated in such a way that a digital model of a traffic corridor is generated using current traffic conditions detected by at least one road traffic sensor, a traffic flow is simulated based on the digital model, and driving behavior instructions are determined based on the simulation of the traffic flow.

[0063] In the situation shown, the vehicles in cell 828 are in a traffic jam and are temporarily stopped. Some of these vehicles are configured as networked motor vehicles 834 and report their own position and detected surrounding objects to the infrastructure unit 804 using CAM and / or CPM messages 814. From this, the infrastructure unit 804 determines the position of the end point of the traffic jam and the risk of potentially occurring braking shock waves. The vehicles in cells 822 and 824 are approaching the traffic delay and cannot yet detect it with their own sensors, either due to the distance or due to a field of view occlusion in dense traffic.

[0064] The infrastructure system 812 (comprising the infrastructure units 804, 808 and the traffic control unit 806) can detect the suboptimal speed conditions of the oncoming traffic and disseminate corresponding driving behavior advice 816, which, for example, aims to progressively reduce the speed of the vehicles 832 in the sections 822, 824 the closer they are to the end point of the traffic jam. For each of the sections 822, 824, 826, driving behavior advice 816 is generated, comprising, for example, a maximum recommended speed, a preferred lane, lane change recommendations, and / or instructions for yielding right of way. If necessary, recommendations can be further divided depending on a vehicle category; for example, a recommended maximum speed for trucks can differ and be defined more specifically.The driving behavior instructions 816 can then be transmitted by one or both of the infrastructure units 804, 808; preferably, one message 816 is sent per section 822, 824, 826, 828.

[0065] Some of the vehicles 834 in section 828 are networked and use ACC or UX devices. Here, these received driving behavior instructions 816 can be implemented via a human driver or a driver assistance system. These vehicles thus influence the possibly non-networked surrounding traffic through their own behavior. Thus, the entire traffic can be indirectly influenced by the infrastructure. The transition between oncoming traffic and congestion can thus become smoother, and speed differences are reduced accordingly, which can increase safety and driving comfort. A lower maximum speed can also contribute to increasing the overall traffic flow on average. QUOTES CONTAINED IN THE DESCRIPTION

[0000] This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions. Cited patent literature

[0000] DE 10 2020 102 329 A1

[0005] DE 10 2010 050 805 A1

[0006]

Claims

[1] Method for traffic flow control, wherein driving behavior instructions (816) are generated for at least one road user, in particular a networked motor vehicle (112, 114, 116, 832, 834, 835) and are sent to the at least one road user, in particular to the networked motor vehicle (112, 114, 116, 832, 834, 835), characterized by that the driving behavior instructions (816) are generated in such a way that a digital model of a traffic corridor (810) is generated by means of current traffic conditions and a traffic flow is simulated based on the digital model and the driving behavior instructions (816) are determined based on the simulation of the traffic flow. [2] Method according to claim 1, wherein data (702) of an infrastructure-side sensor system (102) are acquired for the digital model of the traffic corridor (810), wherein the data (702) represent current traffic conditions and the data are processed together with topological information (704) of the traffic corridor. [3] The method according to claim 2, wherein the topological information of the traffic corridor is comprised of a digital map of a road network. [4] Method according to claim 2, wherein the topological information of the traffic corridor is determined from perception data (814) received from a plurality of networked motor vehicles, in particular by means of machine learning. [5] Method according to one of claims 1 to 4, wherein the simulation is designed as a real-time predictive model with a rule-based traffic control. [6] The method of claim 5, wherein the simulation implements a machine learning approach, wherein the machine learning model is trained with historically recorded traffic conditions. [7] Method according to claim 5, wherein the simulation implements a physical modeling of the traffic flow, in particular a numerical simulation. [8] Method according to claim 5, wherein the simulation implements a machine learning approach in combination with a physical modeling of the traffic flow, in particular a numerical simulation. [9] Method according to one of claims 5 to 8, wherein by means of the real-time predictive model for specific sections (112, 114, 116, 832, 834, 835) of the traffic corridor (810) an optimal driving behavior is determined and driving behavior instructions (816) are derived therefrom, wherein the driving behavior instructions (816) include in particular speed recommendations and / or acceleration and deceleration targets, and / or recommendations for distances to be maintained from the vehicle in front and / or recommendations for the use of a specific lane [10] Method according to one of the preceding claims, wherein the driving behavior instructions (816) comprise at least speed recommendations. [11] Method according to one of the preceding claims, wherein the driving behavior instructions (816) are calculated and transmitted specifically for a particular lane. [12] Method according to one of the preceding claims, wherein the simulation comprises a self-learning algorithm, wherein a continuous optimization of the traffic flow control takes place during operation. [13] Method according to one of the preceding claims, wherein the driving behavior instructions are transmitted to a plurality of road users by means of a multicast and / or a broadcast communication method. [14] Method according to one of the preceding claims, wherein the driving behavior instructions are received by a networked motor vehicle (112, 114, 116, 832, 834, 835) and are output to the driver of the motor vehicle, in particular audio-visually and / or haptically. [15] Traffic control unit (106, 706, 806, 906) designed to generate driving behavior instructions (712, 816) for at least one road user, in particular a networked motor vehicle (112, 114, 116, 832, 834, 835), wherein the driving behavior instructions (712, 816) are generated in such a way that a digital model of a traffic corridor (810) is generated by means of current traffic conditions and a traffic flow is simulated based on the digital model and driving behavior instructions (712, 816) are determined based on the simulation of the traffic flow. [16] Device for traffic flow control, in particular infrastructure system (812) for driving support of at least partially automated networked motor vehicles (832, 834, 835), and in particular designed to carry out a method according to one of claims 1 to 11, comprising at least one communication unit, at least one traffic control unit (106, 706, 806, 906) according to claim 15 and at least one road traffic sensor (102), wherein the traffic control unit (106, 706, 806, 906) is designed to generate driving behavior instructions (712, 816) for at least one road user, in particular a networked motor vehicle (112, 114, 116, 832, 834, 835), and wherein the communication unit is designed to transmit the driving behavior instructions (712, 816) to at least one road user, in particular to a networked motor vehicle (112, 114, 116, 832, 834, 835),wherein the current traffic conditions are detected in particular by means of the road traffic sensor (102).

Images