Research On Motion And Algorithm Of Humanoid Robot Driven By Depth Camera Data

Robot technology has always been a hot spot in the field of artificial intelligence.In order to improve social productivity and improve people's lifestyles,generations of researchers are committed to the study of robotics in order to liberate human labor or special occasions instead of humans.With the continuous improvement of theoretical knowledge and practical experience,various types of robots began to appear in people's daily lives.Of course,the most common ones are the control of the robot based on the controller or voice,such as some wheeled movement is more popular,and walking like humans is rare.Because the knowledge involved in walking is more complicated and it is not easy to design the product,this paper uses the humanoid robot as the research object.Compared to traditional controllers and voice-controlled robots,this paper will use the data collected by Kinect to control the robot using gesture and hand recognition.The new human-computer interaction method is combined with the popular humanoid robot to complete the research of this topic.The main research contents are as follows:(1)After learning more about the working principle of the Kinect sensor,it uses the color data,depth data and skeleton data collected for gesture recognition and hand recognition.Through the skeleton data collected by Kinect,the spatial data obtained by characterizing the joint depth data(spatial information)is used to calculate the result according to the angle of the space joint vector representing the motion,thereby completing the recognition and judgment of the human body posture,and performing the posture recognition experiment and discovering Real-time human body gesture recognition rate is not bad.(2)Compared with the traditional hand-shaped recognition with hand-type original image and hand-shaped contour map as data features,this paper proposes a hand-shaped recognition method based on hand-shaped skeleton as a data feature,and finds that its recognition rate has a very large improvement.In order to meet the ability to identify the hand at a long distance,this paper proposes the data collected by Kinect,combined with the color image and depth image of the hand,to perform related image processing,obtain a clear hand shape,and then perform skeleton formation.In order to obtain its recognition performance,relevant experiments were designed for analysis and verification.Using the Arduino Mega 2560 controller,assemble the humanoid robot as the lower computer.In order to complete the walking algorithm of the humanoid robot,the kinematics model is first established by using the positive motion method.In order to make the humanoid robot stable walking,the trajectory of the ankle joint movement of the swinging leg should have a second-order smoothness,that is,its trajectory function can be guided and continuous in the second order.Therefore,the cubic spline interpolation curve is selected as the trajectory function,and the angle of each joint is solved by inverse kinematics analysis,thereby compiling the joint angle sequence.In order to avoid the failure of the steering gear due to overload during the experiment,stability verification was also proposed in the experiment,which improved the survival rate of the steering gear.(3)Using the OpenCV library,build the human body posture and hand recognition program on the Visual Studio 2013 platform,and write the motion algorithm of the humanoid robot on the Arduino IDE.The other motion algorithms are uploaded to the controller of the humanoid robot,and serial communication is used.The whole system design and implementation of the walking and other movements of the humanoid robot are controlled by the recognition of the human body posture and the hand shape.