Keywords

1 Introduction

In the context of urban smart mobility networks, it is essential for vehicles to establish communication links with one another, the surrounding infrastructure, and fellow traffic participants. Through the implementation of Vehicle2X (V2X) communication, the exchange of crucial data such as the vehicles’ positions, driving dynamics and intentions becomes feasible and yield the extension of the complexity of automated driving functions, such as cooperative driving in complex urban scenarios [1]. This paper utilizes a methodology for the cooperative driving of automated vehicles in mixed traffic scenarios. This methodology applies V2X-communcation and a graph-based cooperation algorithm to optimize urban cooperative intersection scenarios with regard to overall time and energy efficiency [2]. The resulting vehicle-individual trajectories can be further utilized by optimizing dedicated battery electric vehicles (BEV) drive systems with consideration of the automated driving system [3]. The impact of automated driving systems on the layout of drive systems and fuel and energy consumption was analyzed previously [4, 5]. Within this paper, the impact of connected driving on the dimensioning of the drivetrain and the energy consumption is analyzed in comparison to customer operation.

2 Methodology and Approach

The urban traffic scenarios analyzed in this paper are representative excerpts from an urban traffic simulation applied to the research intersection in Braunschweig. These scenarios are derived from real world measurements by a test vehicle equipped with LIDARs. The scenarios involve various (BEV) vehicles in a complex urban environment, as shown in Fig. 1.

Fig. 1.
figure 1

Research intersection in Braunschweig with marked traffic rules and the resulting SUMO Simulation

Full size image

These vehicles form a cooperative network with each other. The trivial approach of solving the conflicts inherent in such a network are traffic lights with fixed time control. Alternatively, in the absence of traffic lights, a trivial “first come first serve” approach, where every vehicle is driving in the order it arrived at the intersection is applied. Both these approaches can be optimized with regard to overall time and energy efficiency by taking advantage of the possibility to communicate within this network. Therefore, the real world traffic scenarios are transferred to a graph based structure in order to optimize the cooperative network. Hereby, the graph is defined by a fixed set of rules:

  • Rule 1: Every traffic participant is a node in the graph.

  • Rule 2: Every traffic participant has a characteristic driving maneuver after whose execution the traffic participant is no longer relevant for the situation and vice-versa.

  • Rule 3: Every traffic participant’s characteristic maneuver may block and may be blocked by another traffic participant. The execution time after which participant A clears the path for traffic participant B is the weight of the directed edge connecting the two corresponding nodes.

  • Rule 4: If a traffic participant’s characteristic maneuver does not block another traffic participant, the weight of the connecting edge is 0.

These rules can be applied to any traffic scenario, resulting in a graph with a corresponding adjacency matrix, as shown exemplarily in Fig. 2. By formalizing the cooperative networks in a graph structure, it is possible to optimize the driving order of the vehicles by interpreting it as a traveling salesman problem. By visiting every node of the graph exactly once, a unique optimized driving order is calculated. Additionally, it is possible to define a set of prioritization rules, granting right of way to public transport, goods transportation or emergency vehicles by visiting these nodes first. According to the driving order calculated by this methodology, the vehicles adapt the speed curve of their individual trajectories according to a pre-defined safety distance of 2 s. Even though this approach can be calculated decentralized, this paper focuses on the calculation by a centralized cooperation infrastructure.

Fig. 2.
figure 2

Transfer of driving scenario to graph structure and adjacency matrix

Full size image

By comparing the trajectories of cooperative networks with different vehicle densities in both conventional fixed time traffic light control and the cooperation optimization algorithms, the impact of the control algorithm on the energy consumption is estimated. As shown in Fig. 3, the resulting trajectories are optimized towards the energy consumption and drivetrain topologies of the connected vehicles, further optimizing the overall network regarding its efficiency. The methodology is explained in detail in another AVEC paper.

Fig. 3.
figure 3

Overall methodology with scenario generation & optimization, requirement engineering und drivetrain optimization

Full size image

The simulated scenarios contain 6 D-segment vehicles with the trajectories and vehicle parameters shown in Fig. 4. For the V2X-hardware and auxiliary components, an additional consumption of 15 W and mass of 2 kg is assumed for each vehicle. In total, these vehicles are applied to 7 scenarios in 3 groups. First and second, the fixed time traffic light control (FTTL) and the “first come, first serve” approach, each with safety distances of 0, 1 and 2 s. Third, the cooperative algorithm by graph based optimization described beforehand. In both the FCFS scenarios and the cooperative scenario, the vehicles are connected with each other and the intersection via V2X.

Fig. 4.
figure 4

Vehicle parameters and trajectories in the simulated scenarios

Full size image

3 Result Discussion and Outlook

The results of the simulation toolchain are compared regarding their overall improvement towards time and energy consumption in an average traffic scenario with 6 vehicles. Hereby, the cooperative driving algorithm with a safety distance of 2 s between the vehicles is compared to a fixed time traffic control (FTTL) and the trivial “first come first serve” (FCFS) approach, each with safety distances of 2, 1 and 0 s. A safety distance of 0 s means, that there is no safety distance at all, the vehicles are driving immediately after each other. The metric to evaluate the overall time efficiency of the cooperative driving maneuver is the overall execution time, which is the sum of the times each vehicle takes to finish the scenario. The results shown in Fig. 5 indicate, that there is only slight to no gain in time efficiency when decreasing the safety distances for either fixed time traffic control and “first come first serve”, with the latter being slightly more efficient overall. The cooperative algorithm increases the overall time efficiency, as previously shown in [2]. When applied to these urban scenarios, the cooperation algorithm optimizes the overall execution time of the scenario by 38% compared to the conventional fixed time traffic light control and by 15% compared to the “first come first serve” approach, while simultaneously keeping the same safety distance between the vehicles.

Fig. 5.
figure 5

Summed overall execution time of each scenario

Full size image

When analyzing the optimized trajectory and drivetrain adaption with regard to the resulting overall energetic optimization, Fig. 6 shows the gain in efficiency when comparing two different drivetrain topologies (1 & 2 speed transmission BEV) in each driving scenario. Hereby, the drive system topology has no significant impact on the energy demand, mainly due to the short scenarios and the low velocity profiles with a high share of acceleration operations. Aside from the drivetrain topology, the overall energy consumption decreases as the total scenario execution time decreases, mainly because of the reduced energy consumption of the auxiliary consumers and the reduced conversion losses in the drivetrain due to fewer driving operations. This results in a reduction in energy demand of 14% when comparing fixed time traffic light control with the cooperation algorithm and 7% in comparison to the “first come first serve” approach while keeping the same safety distances. Figure 6 shows, that in order to be comparable to the cooperation algorithm with regard to overall energy consumption, FTTL would need to be applied with no safety distances at all, thus making in impractical in real traffic applications.

Fig. 6.
figure 6

Overall energy consumption of each scenario with optimized drivetrains, each with 1 and 2 speed transmissions (1 gear/2 gear)

Full size image

4 Conclusion and Outlook

Within this paper, we investigated the impact of cooperative driving on the example of urban intersections with regard to the overall time and energy efficiency. Therefore, based on measurements in real world traffic, seven different intersection scenarios were generated, ranging from fixed time traffic light control, the original measurements were based on, to varying complexities of cooperative algorithms. Based on the resulting swarm trajectories, a drive system simulation of 1 and 2 speed transmission BEV concepts was applied. Hereby, the drive system topology has no significant impact on the energy demand, mainly due to the short scenarios and the low velocity profiles with a high share of continuous accelerations. When looking at the overall energy consumption, this investigation showed, that cooperative functions enable energy consumptions reduction of up to 14% in such urban scenarios, while simultaneously optimizing the overall time efficiency up to 38%.

For future research, there are various potentials to optimize the energy and time efficiency of such multi vehicle systems further. By applying multi-criteria optimization, the overall scenario can be optimized with regard to both overall execution and time and energy efficiency. Furthermore, expanding the scenarios to a larger inner-urban scale and optimizing the drive systems within each scenario in order to identify the optimal drive system parameters for the overall lowest energy consumption are future expansions of the presented method to investigate additional potentials.