Multi-modal Distributed Simulation Combining Cars, Bicyclists, and Pedestrians
Driving, bicycling, and pedestrian simulators have proven to be valuable tools for investigating the underlying causes of crashes and testing various engineering and education countermeasures designed to reduce the risk of crashes. However, most research efforts have focused on studying how individual drivers, bicyclists or pedestrians perform in a specific scenario and participants are exposed to only scripted (static or dynamic) actors; little has been done to study the interactions of drivers, bicyclists, and pedestrians in controlled simulator experiments. We propose to develop a connected simulator platform that could be used to link vehicle, bicycling, and pedestrian simulators so that they share the same virtual environment in real-time. The proposed project will develop a synchronous and connected distributed platform for studying the dynamic interactions between users as they share a virtual roadway environment and support research aimed at gaining insight into how a variety of road users respond to one another in realistic contexts. The proposed platform will also allow for the connection of external add-on devices including advanced assistance systems such as collision warning and detection systems to the simulation in order to study their design and assess the impacts of such systems on human behavior.
By connecting a group of human-in-the-loop simulators, multiple drivers, bicyclists, and pedestrians can share the same virtual environment and be placed in complex and hazardous conditions that cannot be studied in the real world due to the difficulty of creating the necessary controlled conditions and due to the ethical impossibility of placing subject participants in danger. Also noteworthy is that, the proposed improvement in simulation capabilities allows for greater ecological validity, as these simulators are located in different institutions allowing for the simultaneous, synchronous simulation and measurement of human behavior in connected environments involving users from a more representative participant sample as opposed to the convenience test samples typically available in solo simulator experiments. Using the proposed platform, specific contextual scenarios could be created that test the limits of human behavior and provide a basis for examining how drivers, bicyclists, and pedestrians interact as they respond to safety-critical events. We can specifically simulate and measure human performance in multiple conflict situations (for example: T-bone conflicts, left turn across path conflicts, bicyclist right hook crashes etc.) that are otherwise difficult to truly understand, especially when examining the underlying factors influencing the decision-making process of humans in the immediate second or two preceding an impending collision (Hancock and Ridder, 2003). Since the platform will support the connection of external devices, data generated in real-time could be used to test new advanced assistance technologies designed to detect imminent collisions. The external device connectivity could provide a development and test platform designed to improve the accuracy of in-vehicle and roadside safety technology.
One of the challenges for creating the proposed test platform is the definition and creation of an overall simulation architecture that can integrate simulators built on different software platforms and that have different expectations and requirements for input data and communications interfaces. For example, the data will need to be shared in a form that is understood by all entities, regardless of whether or not the entities are simulators. Another difficulty relates to the ability to simulate situations involving multiple users such that meaningful data can be collected from all entities involved in the distributed simulation. This project will need to develop the procedures required for making such data recording and sharing a possibility. The inherent networked nature of the proposed platform will introduce challenges related to latency and jitter that will need to be studied and taken into consideration when defining the strengths and limitations of the proposed platform.
This project brings together a strong team with extensive experience in building and operating a variety of human-in-the-loop driving and pedestrian simulators. The UW-Madison team has experience with the development of algorithms for analyzing vehicle movements and data generated from roadside infrastructure during hazardous conditions. UMass has extensive experience with the development and integration of real-world road geometries into multiple driving simulator platforms. UI is home to state-of-the-art bicycling and pedestrian simulators and numerous driving simulators, including the National Advanced Driving Simulator. This project will expand previous work by the research team in the integration of multi-modal simulators, simulators built on different software platforms, and external entities not typically considered in the simulation world. The project will include pilot testing with multiple users and a variety of simulators to evaluate the effectiveness of the proposed platform.