Learning robust driving policies without online exploration

Abstract

We propose a multi-time-scale predictive representation learning method to efficiently learn robust driving policies in an offline manner that generalize well to novel road geometries, and damaged and distracting lane conditions which are not covered in the offline training data. We show that our proposed representation learning method can be applied easily in an offline (batch) reinforcement learning setting demonstrating the ability to generalize well and efficiently under novel conditions compared to standard batch RL methods. Our proposed method utilizes training data collected entirely offline in the real-world which removes the need of intensive online explorations that impede applying deep reinforcement learning on real-world robot training. Various experiments were conducted in both simulator and real-world scenarios for the purpose of evaluation and analysis of our proposed claims.

Accepted in ICRA 2021. Due to format limitations of ICRA, we include appendix of our detailed evaluation results in this full version. arXiv admin note: substantial text overlap with arXiv:2006.15110

References in corpus (6)

• Solving Rubik's Cube with a Robot Hand
• Challenges of Real-World Reinforcement Learning
• Hybrid Reward Architecture for Reinforcement Learning
• Benchmarking Batch Deep Reinforcement Learning Algorithms
• Deep Steering: Learning End-to-End Driving Model from Spatial and Temporal Visual Cues
• Investigating Generalisation in Continuous Deep Reinforcement Learning