The pursuit of fully automated driving has been a significant point of interest among researchers and industrial communities alike in recent years. As instantly replacing every vehicle with its automated counterpart is implausible, autonomous vehicles need to operate alongside human driven ones. Human drivers tend to show a lot of variation in their driving behaviour, and making non-optimal decisions is a frequent practice. This chaotic environment makes it difficult for controllers on-board automated vehicles to make optimal decisions. Given advanced control techniques, such as model predictive controllers, that can make use of a valid prediction of other traffic participants' behaviours for a significant performance boost, a method to successfully make such predictions will become appealing. In this paper, considering a host connected and automated vehicle or CAV, the performance of a recurrent neural network is investigated for this task using some standard driving cycles data to predict the velocity of the preceding vehicle for multiple horizons in urban driving scenarios.
In this paper, a hybrid approach for situational awareness in roundabouts is presented that can produce traffic participants' behaviour for arbitrary horizons. This real-time implementable strategy consists of dynamic Bayesian network and a continuous variable prediction module (CVPM) as its subparts, making it a data-driven approach while providing the facility to incorporate experts' knowledge into the predictions. Being a data-driven approach, the data is obtained using SUMO as a simulation platform, and three different CVPMs are experimented with, namely recurrent neural network (RNN), gated recurrent unit (GRU), and long short-term memory networks (LSTM). The chosen RNN yields a correlation higher than 0.895 and RMSE less than 0.036 for 10 seconds predictions.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.