首页>
外国专利>
ONLINE LEARNING METHOD AND VEHICLE CONTROL METHOD BASED ON REINFORCEMENT LEARNING WITHOUT ACTIVE SEARCH
ONLINE LEARNING METHOD AND VEHICLE CONTROL METHOD BASED ON REINFORCEMENT LEARNING WITHOUT ACTIVE SEARCH
展开▼
机译:基于主动学习的基于强化学习的在线学习方法和车辆控制方法
展开▼
页面导航
摘要
著录项
相似文献
摘要
PROBLEM TO BE SOLVED: To provide a computer execution type method for adaptively controlling an autonomous operation of a vehicle.SOLUTION: A critic network in a computer processing system configured so as to autonomously control a vehicle has the steps of: determining an estimated average cost and an approximated arrival cost function which generates a minimum value for an arrival cost of a vehicle when applied by an actor network by using a sample of data passively collected and a state cost; and determining a control input which generates a minimum value for the arrival cost by being applied to the vehicle in the actor network operatively connected with respect to the critic network. The actor network determines the control input by estimating a noise level, using the average cost, the arrival cost determined from the approximated arrival cost function, a dynamic value for control for a current state of the vehicle and the passively collected data.SELECTED DRAWING: Figure 3
展开▼