Social Value Orientation (SVO) is a measure of ones’ preference to allocate rewards between themselves and another person. Previous research in human modeling for autonomous driving has incorporated SVO estimation into their prediction models, however, SVO does not always dictate behavior. The environment itself can influence behavior and even help to prime SVOs. The research presented attempts to integrate the relationship between SVO and the environment into human intent inference as prior knowledge of the system. The idea is to create a better model of human driving that considers drivers’ perception of the environment.
By nature, humans can be very unpredictable with their actions, which makes it difficult to create a perfect Theory of Mind (ToM) model to attempt to predict those actions. Having such a model is extremely important to an autonomous car’s motion planning for safe and efficient interaction with other vehicles. The researcher’s work addresses this uncertainty by introducing a Bayesian confidence value within the model to predict the location of other vehicles with a probability distribution. The model will be validated using the Berkeley INTERACTION dataset. The long-term goal is to create a ToM model with mutual intent inference.