Learning an Optimal Race Driver Policy using TORCS
ثبت نشده
چکیده
According to [1] Americans drive approximately 3 billion miles per year. One way to increase the productivity of our society as a whole is to increase the speed with which people reach their intended destinations. Having an autonomous vehicle capable of high speed travel opens up new possibilities for fast travel by car. Control of a car at high speeds is a hard problem which has sparked competitions for creating the fastest autonomous driver in an open source race car simulation called The Open Racing Car Simulator (TORCS)[8]. Many autonomous drivers exist, however most of them use knowledge of the whole track and other parameters that are not available to a driver in a real world. Furthermore, near optimal controllers for robotics applications have been learned using reinforcement learning in the past [5]. This project will show that it is feasible to learn a driving policy that allows a race car to travel quickly, but that the state representation for the learner needs to be carefully designed.
منابع مشابه
Meta learning Framework for Automated Driving
The success of automated driving deployment is highly depending on the ability to develop an efficient and safe driving policy. The problem is well formulated under the framework of optimal control as a cost optimization problem. Model based solutions using traditional planning are efficient, but require the knowledge of the environment model. On the other hand, model free solutions suffer samp...
متن کاملImitation learning of car driving skills with decision trees and random forests
Machine learning is an appealing and useful approach to creating vehicle control algorithms, both for simulated and real vehicles. One common learning scenario that is often possible to apply is learning by imitation, in which the behavior of an exemplary driver provides training instances for a supervised learning algorithm. This article follows this approach in the domain of simulated car rac...
متن کاملLearning to Repeat: Fine Grained Action Repetition for Deep Reinforcement Learning
Reinforcement Learning algorithms can learn complex behavioral patterns for sequential decision making tasks wherein an agent interacts with an environment and acquires feedback in the form of rewards sampled from it. Traditionally, such algorithms make decisions, i.e., select actions to execute, at every single time step of the agent-environment interactions. In this paper, we propose a novel ...
متن کاملDriving Faster Than a Human Player
TORCS car racing bots have improved significantly in the last years. We show that accurate curvature information for the upcoming corners enables offline learning a near-optimal driving style that consistently beats an expert human player (and the fastest currently known bots). Generalization to other tracks often, but not always succeeds, so that the method is extended by an online error corre...
متن کاملEvolving Drivers for TORCS using On-Line Neuroevolution
We applied on-line neuroevolution to evolve non-player characters for The Open Racing Car Simulator. While previous approaches allowed on-line learning with performance improvements during each generation, our approach enables a finer grained on-line learning with performance improvements within each lap. We tested our approach on three tracks using two methods of online neuroevolution (NEAT an...
متن کامل