A SUMO Extension for Norm-Based Traffic Control Systems

Abstract

Autonomous vehicles will most likely participate in traffic in the near future. The advent of autonomous vehicles allows us to explore innovative ideas for traffic control such as norm-based traffic control. A norm is a violable rule that describes correct behavior. Norm-based traffic controllers monitor traffic and effectuate sanctions in case vehicles violate norms. In this paper, we present an extension of SUMO that enables the user to apply norm-based traffic controllers to traffic simulations. In our extension, named TrafficMAS, vehicles are capable of making an autonomous decision on whether to comply with norms. We provide a description of the extension, a summary on its implementation and demonstrative experiments.

Appendices

Appendix 1: Merge Norm Scheme

For the merge norm scheme, we use the same pseudocode structure (Algorithm 2) as for the stay-on-lane norm scheme. As with the other norm scheme we begin with the instance of the norm (lines 0–8).7 Initially we read sensors 1 and 3 and merge the readings using the algorithm of Wang et al. [21] (line 1). The result is an ordered list of agents, which, if they continue as they are, will arrive at the merge point in the same order. We maintain a global variable $t_{free}$ that indicates the next moment in time that the merge point is free. With optimalVelocity we calculate the optimal speed for an agent s.t. it will arrive at $t_{free}$ on the merge point plus some safe margin, or later if the agent cannot make it in time physically (line 3). If the agent is at the right lane of the main road and the optimal velocity is below a predefined threshold, then it is obliged to move to the left lane (line 5), otherwise it is obliged to adapt its velocity to the optimal velocity and pass the merge point on the right lane (line 7). An agent is sanctioned if it is not passing the merge point on the correct lane (lines 11–12, and 15–16). Otherwise, an agent can also be sanctioned if it did not achieve its predetermined velocity (lines 19–20).

Appendix 2: Using Our Code

Our framework is open-source and available on-line on Github at https://github.com/baumfalk/TrafficMAS. It can be compiled from source, or it can be downloaded as a binary version.

1.1 How to Run It

Our framework can be run as follows. Assuming you use the binary JAR file, a scenario can be run with the following command:

java -jar TrafficMAS.jar ./scen/ scenario.mas.xml path/to/sumo scenario.sumocfg [seed].

In this command scen is the directory the scenario is located in, scenario.mas.xml is the main configuration file for the scenario and path/to/sumo denotes the SUMO executable to use. The SUMO-GUI program can also be used. The parameter scenario.sumocfg denotes the SUMO configuration file used by the scenario. Finally, the parameter seed is used to prepare the random number generator, which is used to spawn vehicles in a probabilistic fashion. If no seed is provided, a random one is generated by the system.

1.2 How to Create Your Own Scenario

Our framework also allows for the creation of your own scenarios. A TrafficMAS scenario consists of several XML files:

• a global configuration file, containing the paths to the other XML files, as well as the simulation duration.

• a configuration file specifying which norms are used. In this file, the norms are also parameterized with scenario-specific information, such as road names.

• a configuration file which describes the norm-based traffic controllers. The file is used to define which controllers there are, which sensors they have access to and to which other controllers they are subscribed.

• a configuration file containing the vehicle profile distributions. This file contains the distributions of the various driver profiles and the traffic density of the different roads.

• various SUMO XML files: the XML file containing the nodes, the edges, the sensors and the routes.