SuperDriverAI: Towards Design and Implementation for End-to-End Learning-based Autonomous Driving

Shunsuke Aoki,Issei Yamamoto,Daiki Shiotsuka,Yuichi Inoue,Kento Tokuhiro,Keita Miwa

2023-05-14

Abstract:Fully autonomous driving has been widely studied and is becoming increasingly feasible. However, such autonomous driving has yet to be achieved on public roads, because of various uncertainties due to surrounding human drivers and pedestrians. In this paper, we present an end-to-end learningbased autonomous driving system named SuperDriver AI, where Deep Neural Networks (DNNs) learn the driving actions and policies from the experienced human drivers and determine the driving maneuvers to take while guaranteeing road safety. In addition, to improve robustness and interpretability, we present a slit model and a visual attention module. We build a datacollection system and emulator with real-world hardware, and we also test the SuperDriver AI system with real-world driving scenarios. Finally, we have collected 150 runs for one driving scenario in Tokyo, Japan, and have shown the demonstration of SuperDriver AI with the real-world vehicle.

Robotics,Artificial Intelligence

What problem does this paper attempt to address?

The paper aims to address the issue of achieving fully autonomous driving for self-driving cars on public roads. Specifically, despite the significant progress made in current autonomous driving technology, there are still two major challenges: 1. How to effectively collaborate with surrounding pedestrians and other human drivers; 2. How to handle unexpected road scenarios caused by construction, traffic control, obstacles, or potholes. To solve these problems, the paper proposes an end-to-end learning-based autonomous driving system called SuperDriver AI. This system uses deep neural networks (DNN) to learn driving behaviors and strategies from experienced driver data and determines driving operations while ensuring road safety. In addition, to improve the robustness and interpretability of the system, the paper also introduces the slit model and visual attention module. By constructing a data collection and simulation system based on real hardware and vehicles, the researchers tested SuperDriver AI and collected 150 runs of data in a driving scenario in Tokyo, Japan, demonstrating the practical application of the system. Overall, the main contribution of the paper is the proposal of an end-to-end learning-based autonomous driving system and the demonstration of its feasibility and effectiveness in the real world.

SuperDriverAI: Towards Design and Implementation for End-to-End Learning-based Autonomous Driving

Autonomous Vehicle Control: End-to-end Learning in Simulated Urban Environments

End-to-End Autonomous Driving Through Dueling Double Deep Q-Network

Deep Imitation Learning for Autonomous Driving in Generic Urban Scenarios with Enhanced Safety

Deep Learning for Autonomous Driving System

End-to-end deep learning-based autonomous driving control for high-speed environment

End-to-End Autonomous Driving With Semantic Depth Cloud Mapping and Multi-Agent

Humanlike Driving: Empirical Decision-Making System for Autonomous Vehicles

Deep learning and control algorithms of direct perception for autonomous driving

Efficient and Generalized End-to-end Autonomous Driving System with Latent Deep Reinforcement Learning and Demonstrations

DeepTest: Automated Testing of Deep-Neural-Network-driven Autonomous Cars

Recent Advancements in End-to-End Autonomous Driving using Deep Learning: A Survey

End-to-End Deep Reinforcement Learning for Lane Keeping Assist

Safe and Generalized end-to-end Autonomous Driving System with Reinforcement Learning and Demonstrations

End-to-End Urban Autonomous Driving With Safety Constraints

Learning Robust Control Policies for End-to-End Autonomous Driving From Data-Driven Simulation

An End-to-End Driver Simulator for Personal In-Vehicle Conversational Assistant

Multimodal End-to-End Autonomous Driving

Multi-Modal Sensor Fusion-Based Deep Neural Network for End-to-End Autonomous Driving With Scene Understanding