BEHAVE: Behaviour Evaluation of Human-driven and Autonomous VEhicles
This project is a collaboration between TUMCREATE and Berkeley, funded under the NRF seed grant scheme.
Inevitably the introduction of autonomous vehicles (AVs) will have far-reaching effects on traffic conditions. Experimenting in real traffic conditions is obviously an expensive and potentially unsafe proposition. To mitigate this problem, a great deal of preliminary analyses are carried out virtually, i.e., using simulation techniques to evaluate various what-if scenarios.
The current state-of-the-art in traffic and mobility models, however, does not allow a satisfactory and realistic evaluation of traffic behaviour when both autonomous and man-driven vehicles are present. The aim of this project is to provide a framework for studying mixed ( AVs/conventional vehicles (CV)) traffic and evaluating implications of a wide range of AV-related policies.
To achieve this aim, we are using the simulation framework offered by CityMoS, an agent-based simulator developed at TUMCREATE. However, CityMoS alone will not provide a solution to mixed traffic planning and regulation unless the following two essential tasks are carried out that constitute the two main milestones of this project:
Modelling human drivers adequately
The behaviour of vehicles in microscopic simulation is described by means of models that determine the actions of the driver. Existing driving models attempt mimicking human behaviour, however, they do not take into account a sometimes seemingly irrational and highly unpredictable behaviour. Furthermore, typically models available in the literature are collision-free. Modelling accidents at a microscopic level is crucial as they are a vital part of the analysis of safe AV integration.
In order to mitigate the gaps between current simulation models and reality this project will model some of the human-typical factors such as attention, perception and aggression as stochastic processes and therefore extend existing models. The high level of flexibility of our approach will allow modelling accidents in our simulation environment, something that is obviously impossible using conventional collision-free models. Furthermore, the extended models will also take into consideration environment effects such as weather conditions, darkness, rush-hour aggression etc.
Modelling autonomous vehicle behaviour including the effects of different AV policies on the system such as specialized AV lanes or regulations regarding platooning.
The AV model should be at the same time generic enough to allow easy extension to more specific AV strategies and yet complex enough to capture the wide spectrum of AV functionality. We propose to develop the models describing autonomous vehicle behaviour based on a hierarchical structure.
At the bottom, there is the "standard" car–following and lane-changing model referred to as Low-Level Control. This model takes as input the description of the immediate surroundings of the agent and determines the autonomous control of the vehicle.
At a higher level of the control hierarchy, we will develop a Trajectory Planner model that provides inputs to the low-level control to take into consideration broader traffic state information. This planner performs the high level planning needed to achieve the goals set by the High-Level Decision Module thus enabling maneuvers such as the one shown in the figure below, which go beyond simple platoon formation, join and leave commands.
Outcomes and Impact
In addition to the evaluation of various AV scenarios, the developed framework can be used to detect potentially unforeseen consequences of AV introduction. Such findings may include identifying driver types that are imposing severe restrictions on safe AV integration, road network settings that are manageable for humans but not for AVs, road infrastructure that is made obsolete by the presence of AVs, and accident triggers that only exist in mixed-traffic conditions.
Our research will open the doors to the predictions of non-trivial emerging behaviours thus allowing researchers and policy makers to plan for a smooth introduction of autonomous vehicles into current transportation systems.
This project is a basis for more far-reaching and deeper investigation of research questions arising from traffic safety, evaluation of different AV driving strategies, and even the optimization for AV-based ride sharing systems.
Furthermore, the extension of this project may, instead of locally optimizing AV trajectories, develop global strategies, e.g., how AVs can be distributed onto different roads to reach a system optimum, not only with respect to travel times but possibly also in terms of charging strategies for electrical vehicles or even heat distribution and noise pollution.