Learning from experience for rapid generation of local car maneuvers. (October 2021)

• Record Type: Journal Article
• Authors: Kicki, Piotr; Gawron, Tomasz; Ćwian, Krzysztof; Ozay, Mete; Skrzypczyński, Piotr
• Is Part Of: Engineering applications of artificial intelligence. Volume 105(2021)
• Publication Date: 2021-10
• Subjects: Motion planning -- Neural networks -- Robotics -- Autonomous driving -- Reinforcement learning -- Autonomous vehicle navigation; Engineering -- Data processing -- Periodicals; Artificial intelligence -- Periodicals; Expert systems (Computer science) -- Periodicals; Ingénierie -- Informatique -- Périodiques; Intelligence artificielle -- Périodiques; Systèmes experts (Informatique) -- Périodiques; Artificial intelligence; Engineering -- Data processing; Expert systems (Computer science); Periodicals; 620.00285
• DOI: 10.1016/j.engappai.2021.104399
• Languages: English
• ISSNs: 0952-1976

Abstract

Abstract: Being able to rapidly respond to the changing scenes and traffic situations by generating feasible local paths is of pivotal importance for car autonomy. We propose to train a deep neural network (DNN) to plan feasible and nearly-optimal paths for kinematically constrained vehicles in a small constant time. Our DNN model is trained using a novel weakly supervised approach and a gradient-based policy search. On real and simulated scenes and a large set of local planning problems, we demonstrate that our approach outperforms the existing planners with respect to the number of successfully completed tasks. While the path generation time is about 40 ms, the generated paths are smooth and comparable to those obtained from conventional path planners. Highlights: Reinforcement learning based path planning for car-like vehicles. Fast generation of local maneuvers for cars using machine learning. A novel differentiable loss function for training path planning neural networks. A dataset of real-world planning problems for comparing path planners. Real-time path planning for autonomous vehicles in the CARLA simulator.