基于替代传导径迹增强式学习的自主式微直升机控制

随着微电子机械系统（MEMS）的迅猛发展，自主式微直升机的研究也已成为这一领域内的研究热点之一．由于微直升机尺寸的限制，不能安装功能很强的传感器和处理器，难以获得完全的环境信息，所以传统的基于模型的控制方法不适用于环境是动态的自主微直升机控制．基于行为的控制方法采用累次逼近的方法，不需要环境的精确模型，因此系统的稳定性较好．本文采用基于替代传导径迹的增强式学习，结合即时差分方法，提高其学习效率，仿真实验验证了该学习算法的有效性．最后，本文介绍了微直升机控制中存在的一些问题和 我们以后的改进方向．

AUTONOMOUS MICRO-HELICOPTER CONTROL BASED ON REINFORCEMENT LEARNING WITH REPLACING ELIGIBILITY TRACES

With the rapid development of MEMS, study of micro helicopter has been a hotpot in this field. Because of its overall size, the micro helicopter could not be equipped with strong sensors and MPU, which affect the helicopter to get the whole environment information, therefore, the traditional control method disagrees with the helicopter in the uncertain environment. However, the method based on behavior only uses trial and error without the exact model of the environment. We adopt reinforcement learning with replacing eligibility traces to be combined with the temporal difference learning, which improves the efficiency and speed convergence. The results of simulation prove the validity of the learning algorithms. At last, this paper introduces the existent problems with the helicopter control and gives the future study trend.