4results about How to "Shrink state space" patented technology

The present invention discloses a bus scheduling method based on reinforcement learning, which belongs to the field of bus scheduling, and specifically comprises the following steps: first, the vehicle scheduling problem of a certain bus line is modeled as a Markov decision process, which includes three elements: state, action, and reward. Each time point in the departure schedule is taken as a decision point, and a reinforcement learning agent is used to select a currently available vehicle to be dispatched from the time point, thereby covering the time point. A simulation environment for bus scheduling is constructed, and the optimal strategy of the agent is obtained by using a reinforcement learning algorithm through the interaction between the agent and the simulation environment. Then, the agent is used to make decisions point by point in time sequence for each time point in the departure schedule, thereby obtaining a bus scheduling plan. The present invention provides an online scheduling method for buses, which can ensure that the departure schedule is fully covered in the event of uncertain events such as traffic congestion and traffic accidents, and a small number of vehicles are used.

The invention provides a natural gas P2 heavy load vehicle-oriented EMS adaptive mixed domain dynamic planning method, and relates to the field of new energy. The invention provides a self-adaptive hybrid domain dynamic planning method. Through systematic layering, the state and action space is greatly reduced on the premise of maintaining the calculation precision. In order to solve the problem that a large-capacity battery is neglected in discretization calculation due to small single-step SOC change, a coarse and fine combined adaptive SOC grid is constructed to dynamically adjust a calculation grid. Three-dimensional fitting is carried out on multiple gears of a motor and an engine based on a physical model so as to establish an optimal gear shifting model, and then the current optimal gear is rapidly locked according to the accelerator opening degree and the vehicle speed. And the optimal motor-engine torque combination is quickly screened in a mode of combining gradient descent and global search. By adopting the AHD-DP method disclosed by the invention, excellent robustness, optimality and generalization ability are shown.

The present application relates to the technical field of human-computer interaction, and particularly relates to a scenic spot robot service system based on AI interaction and unmanned driving technology. The system automatically constructs scenic spot attribute labels by mining internet evaluation texts; obtains user multi-modal demand descriptions and projects them to label semantic space to generate screening conditions and reordering weight vectors; combines robot remaining range to define geographical topological fences, preliminarily screens and filters out a candidate scenic spot set; establishes a dimensionless evaluation function containing normalized actual path cost and estimated tour benefits, and iteratively searches to generate a self-defined tour route by using an optimized path planning algorithm. During driving, the system captures unplanned scenic spots based on voice, vision or motion interaction, dynamically expands the set and reiterates the algorithm after confirming the real intention through active AI inquiry, and realizes adaptive correction of the route. The present application realizes global collaborative optimization of physical space constraints and subjective intention gain.

This invention discloses a radar anti-jamming decision-making method that combines online domain optimization with intra-domain decision-making. It aims to address the problems of low anti-jamming benefit values ​​in traditional radar anti-jamming decision-making methods and their inability to adapt to scenarios with numerous jamming states and a large number of anti-jamming measures. The implementation steps of this invention include: determining the radar anti-jamming decision domain online using template matching technology; generating an anti-jamming benefit value sequence; and generating anti-jamming decision results within the radar anti-jamming decision domain. This invention, while ensuring decision-making performance, directly selects the anti-jamming measure with the highest anti-jamming benefit value sequence as the anti-jamming decision result, greatly improving time efficiency and anti-jamming benefit, and facilitating practical application.