| Mobile robots have been developed very fast recent years.Represented by sweeping robots,indoor mobile robots have been used by many families.However,most of them use lidar to navigate in the environment,which results in the high-cost expense and prevent them from becoming popular.Besides,most researches in China focused on just a particular area of navigation and causes only a small amount of navigation system to exist.Based on this,this thesis does some research on the technology used in robot navigation and designs a vision-based navigation system on a low-cost hardware platform.The main contributions of this thesis are as follows:1.The planning task which indoor mobile robots should complete is discussed and D*Lite algorithm is chosen to be the system's path planning algorithm.The advantages and shortcomings of D*Lite algorithm are analyzed.After that,line of sight algorithm and distance transform is used to overcome the shortcomings.The improved algorithm is implemented and experimental results shows that the algorithm is correct and effective.2.The localization problem of mobile robots is introduced.The formula of dead reckoning is derived and the error of using that is analyzed.Vision based localization is then used to reduce the error and its basic theory is elaborated.The kind of map is determined based on the planning task and the observation models of various sensors are established.Probabilistic method is applied to alleviate the difference between real world and the established map.3.The function requirements of indoor mobile robots are analyzed and function submodules of the system are designed.The navigation system is partitioned into deliberate layer and reactive layer.The perception module and the control module on the reactive layer are designed.The system structure of deliberate layer and the working process of the submodules on the deliberate layer is expounded.4.The requirements of the network module is analyzed.The basic function of server is achieved on the robot and the client is implemented using Qt on the computers.The communication protocol of them is designed and Google Protobuf library is used to simplify the encoding and decoding process.The navigation system is tested on the sweeping robot developed by the laboratory.The sweeping robot uses STM32 and raspberry pi as core processers.Results show that the robot using the system can achieve the goal in unknown environments.The error of localization is about 6cm.On the coverage planning mode of the robot,the coverage percentage is 91.47%.The map built by the robot is almost same as the environment,and the robot can communicate with the computer in normal. |
PDF Full Text Request