What will the marvelous future of “connected vehicles” mean for drivers in the real world? To answer this crucial question, Toyota Motor North America Research and Development is collaborating with the University of Massachusetts to award Associate Professor Daiheng Ni of our Civil and Environmental Engineering Department a four-year research grant so he can explore connected vehicle technology. As Professor Ni explains, “The outcome of this research can serve as the input to better powertrain management and further optimize vehicle control that can potentially transform the way that we drive in the future while helping save lives and fuel.”
The title of Ni’s Toyota research is “Driver Modeling to Predict Operational Control on a Short Horizon.” The overall objective of Ni’s Toyota research is to develop a mathematical model to predict driver’s operational control in terms of anticipated speed and acceleration on a short horizon of between 0 and 10 seconds of reaction time.
Connected vehicles are those that will enable safe, interoperable networked wireless communications among vehicles, the infrastructure, and passengers’ personal communications devices.
In anticipating future challenges while maintaining a leadership position in the era of connected vehicles, Toyota collaborating with the University of Massachusetts to help make driving safer, more fuel efficient, and more environment-friendly in this interconnected context. Toyota is interested in exploring Professor Ni’s pioneering research on “Field Theory in Transportation” so that it may understand how his work may help predict drivers’ responses in a dynamic way within the mix of road traffic, especially in a connected vehicle setting.
As principal investigator on this project, Ni says that his preliminary work on the Field Theory in Transportation is ideal for this project because it is mathematically amenable to representing the natural way of human thinking and reasoning and readily incorporates the modeling of connected vehicle effects.
Basically, in this field theory, physical world objects (e.g. roadways, vehicles, and traffic control devices) are perceived by the subject driver as component fields.
“The driver interacts with an object ‘at a distance,’ as opposed to ‘by contacting,’” explains Ni, “and the interaction is mediated by the field associated with the object. In addition, the field may vary when perceived by different drivers, especially those who are assisted by connected vehicle technology. The superposition of these component fields represents the overall hazard encountered by the subject driver. Hence, the driver’s operational control can be captured as seeking the least hazardous route by navigating his or her vehicle through the overall field.”
Ni explained this research in an article titled “A Unified Perspective on Traffic Flow Theory, Part I: The Field Theory,” in the journal of Applied Mathematical Sciences, vol. 7, no. 39, pp. 1929-1946, 2013.
Ni will take a field theory approach to address an aspect of Toyota’s NEXTCAR project, which seeks to improve fuel economy in anticipation of the era of connected vehicles. Ni’s projects seeks to develop a driver model to predict demands by building on his prior research on driver modeling and microscopic traffic simulation especially his preliminary research on field theory of traffic flow.
“More specifically,” says Ni, “I will investigate the mechanism of drivers’ operational control, based on which to capture such a mechanism using the field theory. With the resultant model, prediction can be made on drivers’ demand/pedal actuation given a set of external conditions, such as the presence of surrounding vehicles, roadway conditions, and traffic control devices.”
Specifications for the driver model will be based on information from the operator’s driving environment, including his or her own vehicle, surrounding vehicles, roadways, traffic control devices, and other transportation facilities via onboard sensors and connected vehicle technology. The proposed driver model should be able to predict appropriate next control strategies such as speed and acceleration in order to maintain safety and mobility in the traffic stream.
The proposed driver model is also expected to mimic real drivers’ control behavior by means of processing external inputs and determining output for vehicle control.
Ni’s research is expected to cast prediction in a short horizon of 0-to-10 seconds of reaction time under certain assumptions on how the driving environment may evolve. With forecasted future conditions, the proposed driver model should be able to predict anticipated control strategies.
The proposed driver model should allow user customization to fit the specifics of a particular driver. Meanwhile, the model should also perform online learning to adapt to a specific driver’s control behavior and parameterize the model when needed. This model is expected to undergo empirical validation by using field data collected from the real world. In model validation, the proposed model will be implemented via computer, by which computer simulation will be performed to examine how simulated driver responses approximate those of real drivers. (January 2018)