Transferring Autonomous Driving Knowledge on Simulated and Real\n Intersections

2017/11/30 by David Isele, Isele, David, Akansel Cosgun +1 · 14 citations
Computer Science · Engineering · Psychology · #Adversarial Robustness in Machine Learning #Artificial Intelligence (cs.AI) #Artificial intelligence #Autonomous Vehicle Technology and Safety #Cognitive psychology #Computer science #Engineering #FOS: Computer and information sciences #Forgetting #Human–computer interaction #Intersection (aeronautics) #Isolation (microbiology) #Machine Learning (cs.LG) #Machine learning #Psychology #Reinforcement Learning in Robotics #Reinforcement learning #Robotics (cs.RO) #Task (project management) #Transport engineering #cs.AI #cs.LG #cs.RO

paper · pdf · doi:10.48550/arxiv.1712.01106

published in arXiv (Cornell University) (Cornell University) · Appeared in Lifelong Learning Workshop @ ICML 2017. arXiv admin note: text overlap with arXiv:1705.01197

arxiv created 2017/11/30 · openalex publication_date 2017/11/30 · arxiv updated 2017/12/05 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

We view intersection handling on autonomous vehicles as a reinforcement\nlearning problem, and study its behavior in a transfer learning setting. We\nshow that a network trained on one type of intersection generally is not able\nto generalize to other intersections. However, a network that is pre-trained on\none intersection and fine-tuned on another performs better on the new task\ncompared to training in isolation. This network also retains knowledge of the\nprior task, even though some forgetting occurs. Finally, we show that the\nbenefits of fine-tuning hold when transferring simulated intersection handling\nknowledge to a real autonomous vehicle.\n

Citations

• Playing Atari with Deep Reinforcement Learning
• Connectionist models of recognition memory: Constraints imposed by learning and forgetting functions.
• An Empirical Investigation of Catastrophic Forgetting in Gradient-Based Neural Networks
• Human-level control through deep reinforcement learning
• A Survey on Transfer Learning
• End to End Learning for Self-Driving Cars
• A Deep Hierarchical Approach to Lifelong Learning in Minecraft
• Safe, Multi-Agent, Reinforcement Learning for Autonomous Driving
• Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World
• CNN Features off-the-shelf: an Astounding Baseline for Recognition
• Analyzing Knowledge Transfer in Deep Q-Networks for Autonomously Handling Multiple Intersections

Cited by

• Predictive calibration for digital sun sensors using sparse submanifold convolutional neural networks
• Strawberry Detection using Mixed Training on Simulated and Real Data

Related

• Analyzing Knowledge Transfer in Deep Q-Networks for Autonomously Handling Multiple Intersections
• Navigating Occluded Intersections with Autonomous Vehicles using Deep Reinforcement Learning
• State Dropout-Based Curriculum Reinforcement Learning for Self-Driving at Unsignalized Intersections
• Driving Tasks Transfer in Deep Reinforcement Learning for Decision-making of Autonomous Vehicles
• Courteous Behavior of Automated Vehicles at Unsignalized Intersections via Reinforcement Learning