摘要
智能体对环境的认识是其进行决策的重要依据。在机器人足球仿真比赛(RoboCup)中,视觉感知是智能体获得信息、构建世界模型的主要途径。针对仿真比赛中球员智能体进行决策时信息不准确的问题,结合优先级和可信值对智能体的视觉策略进行了设计。根据对观察目标需求的紧迫程度动态地指定观察目标的优先级,然后结合世界模型维护的观察目标的可信值生成各个视觉角度的评价值,通过搜索评价值最大的视觉角度获得优化的视觉信息。实验结果表明这种视觉策略使球员动作的执行更加可靠,增强了球队的整体性能。
Environmental cognition is an important basis on making decision for agents . Visual perception is the main way for agents to obtain information and construct the world model . In order to solve the problem of obtaining inaccurate information in Robocup , priorities and confidence are combined to design the visual strategy . The priorities of observation objects are assigned dynamically according to urgency , then evaluation value of each visual angle is generated by combining the confidence which is kept in the World Model , so optimal visual information can be obtained by searching for the visual angle with maximum evaluation value . Experiment results indicate that this visual strategy helps agents execute actions more reliably and the performance of the whole team can be enhanced .
出处
《微型机与应用》
2014年第5期90-92,95,共4页
Microcomputer & Its Applications
基金
扬州市职业大学教科研项目(编号:07202)
