Research On Mobile Robot Exporalation And Path Planning In Unknown Indoor Environment

The mobile robot is an important research orientation in the field of robots and is a multi-discipline combination of robotics, computer science, artificial intelligence, et al. In practical applications, the mobile robot needs to complete the task of exploration and path planning in case of lack of prior knowledge of environment. How to improve the mobile robot’s self-learning ability to implement environment exploration and path planning has become a research hotspot recently and this essay will do some research on this focus.This paper proposes a detection system based on vision to meet mobile robot’s need for target information in unknown environment. This system allows the mobile robot to access target information on the process of path planning. In the system, a target fitting method based on constrained least squares method is provided to improve the accuracy of targeting. In order to increase the contribution of target information, the concept of difference reward was introduced to design the reward signal in the subsequent reinforcement learning.A mobile robot path planning system based on adaptive fuzzy neural-network was designed on the foundation of the mobile robot’s sensors’ detection data, which complete the path planning task by using the environment information. The system combines the fuzzy system’s expression ability and the artificial neural network’s generalization effectively and establishes the relation between environment information and action by training, which demonstrates a good learning ability.This paper provides a reinforcement learning decision-making system based on adaptive fuzzy neural network for the demand of mobile robot’s detection and path planning in the unknown environment. The system allowed the mobile robot learning the relation between the state space and the action space without relying on the model of environment. The rapid expansion problem during Q-function learning can be solved effectively by the adaptive fuzzy neural network’s function approximation capability and generalization. Meanwhile, the system uses the visual information to improve the application value.The proposed methods’ validity has been verified through relevant simulation and practical test. The ideas and results in the research can have some value of reference and application on the development of the mobile robot intelligence to a certain extend.