Scene Recognition And Motion Control Of Amphibious Spherical Robots

Scene recognition and motion control is foundation to realize robot intelligent,such as autonomous navigation,human-computer interaction,the situation judgment.Those are the basis of the robots' system.It is of great significance for the construction of autonomous mobile robot.Thanks to the rapid development of computer science and sensor technology,robots' scene recognition and motion control has attracted lots of researchers' attention.The trend of scene recognition is compound,intelligent,multi-purpose.However,it hasn't been well solved in amphibious complex environment and the small robot application of robots in scene recognition and motion control.This paper focuses on the research on amphibious working environment and the requirements of more complicated task,where there exist some problems in the robot's application,such as unstable and weak processing ability.In order to realize the amphibious switching freely with the stable operation of the amphibious environment and monitoring robot's state,this paper design,implement,and validate a motion control system based on position detection and scene recognition.First,in view of the deficiencies of the core capacity and the unreasonable structure,we upgraded the core controller,redesigned the structure of robot and adjusted underwater seal problems by using the principle of the mechanical,to complete the platform migration with basic previous function of robot.Second,according to the need of robot motion control applications,we analyzed the motion model of robot and the gait,designed and implemented a driving system with multiple speed and multiple gait by using FPGA's programmable and customizability.Third,in order to monitor the robot's state,switch between land and water freely,and climb slope,we designed a motion adjustment module by static measuring the slope angle,with the robot's posture detection system.Fourth,with the use of machine vision and pattern recognition technology,we built a BoF(bags of the features)model for scene recognition,on the basis of extracting SIFT local feature information and constructing scene visual words,and proposed the movement strategy based on scene recognition.With pose detection and scene recognition applied to the robot's movement adjustment,we designed a variety of robot climbing experiments and different scenarios identification experiments to verified the feasibility of the scheme.