Autonomous driving is a rapidly evolving technology for both research and industry due to its ability to provide driver convenience and enhance safety by avoiding traffic accidents. When building planning algorithms for autonomous driving, it is great significance to reason about multiple dynamic agents, traffic rules and social rules in complex semi-structured urban driving environments, which makes the problem difficult. An elegant decision making strategy could be found by leveraging the recent advance of deep learning, which has shown several breakthroughs in artificial intelligence. We aim at developing learning-based autonomous driving algorithms that could handle complex urban autonomous driving scenarios and demonstrate superior performance in terms of safety, comfort, and intelligence.
* Learn state-of-the-art deep learning technologies with the application to autonomous driving
* Help build up a dataset of human driving behaviors, which would be used to neural network training
* Learn to develop the decision-making module for urban autonomous driving using deep learning
* Gain enough experience of building up dataset and labeling the data for deep learning
* Gain enough experience to model human driving behaviors using huge amount of data
* Learn methodologies for decision making of autonomous driving using deep neural networks