| With the development and breakthrough of computer network technology and artificial intelligence technology,low-speed unmanned equipment has developed rapidly.Floor sweeping robots,distribution robots and disinfection robots are widely used in homes,supermarkets,hotels,hospitals and other environments.For mobile robots,building maps is the first step for autonomous navigation and task execution.However,at present,the mapping of robot still depends on remote control and human decision-making.How to make robots explore the environment and complete mapping independently is a key step for robots to move towards intelligence.With the application in indoor environment as the background,this project studies more efficient and intelligent autonomous exploration methods of robots,and improves the exploration efficiency of robots in indoor scenes.First of all,the indoor environment structure is rich in changes,and the open area is easy to exceed the sensing range of the sensor.When exploring the open area,the robot is often unable to effectively update the mapping information,resulting in the robot hovering and low exploration efficiency.In order to solve this problem,this paper proposes an autonomous exploration algorithm based on the income function model of frontiers and the beam model of rangefinder.By estimating the environmental information perceived by the robot in the process of moving to the frontier,and combined with the information gain,environmental information and path cost at the frontier,comprehensively evaluate the action of the robot to reach each frontier,so as to make the robot try to choose the frontier with more abundant environmental information and avoid the wandering problem caused by selecting the frontier with less environmental information in the middle,so as to improve the exploration efficiency of the robot.Secondly,the current exploration strategies mostly rely on the income function model of independent frontiers.In the exploration process,the robot mainly adopts the greedy strategy,selects the frontier with the largest income value at each time,and ignores the geometric continuity of the environment.Before a robot completes a geometrically continuous area(such as a complete room),it often turns to another area,resulting in the disordered movement of the robot.Inspired by the law of bionic human perception,this paper uses the semantic information of the environment as inspiration to guide the robot exploration,and proposes a robot exploration strategy based on heuristic biased-sampling.Firstly,this paper improves the YOLO(you only look once)object detection network and obtains a lightweight model for heuristic object recognition.Secondly,based on the location of the identified heuristic object,a geometric continuous region is constructed to approximately represent the actual room region,and then the frontiers of the region are extracted by the method of fast prime search random tree based on offset sampling.Finally,the heuristic information revenue model is designed to determine the target frontier to be explored.The target point will guide the robot to enter the heuristic area for exploration,and go to other areas to explore after the area is explored completely,which greatly reduces the backtracking phenomenon and improves the exploration efficiency.Finally,in order to verify the feasibility and effectiveness of the algorithm,this paper built a mobile robot experimental platform to verify the proposed autonomous exploration algorithm in the gazebo simulation environment and the actual scene.By changing the complexity of the environment,different experimental scenes are designed to test the performance of the proposed autonomous exploration system in different scenes.The exploration performance of the robot is evaluated from four indicators:the use time of the robot to explore the complete environment,the path length,the speed change curve of map entropy deceleration and the exploration trajectory.The experimental results show that the proposed method can complete the environmental exploration more efficiently,reduce the backtracking phenomenon in the exploration process and improve the exploration efficiency. |
