Vision-based Motion Measurement And Control Of A Soft Robot

Pneumatic soft robots have superior flexibility and suppleness due to their infinite degrees of freedom,and are characterised by high manoeuvrability,simple structure,high safety and environmental adaptability compared to traditional rigid robots,which have broad application prospects in assisting minimally invasive surgery,pipeline inspection,exploration and rescue,underwater grasping,etc.However,when soft robots need to interact with the environment,there are problems such as low motion accuracy and control difficulties.In this paper,a vision-based method for measuring motion parameters,position control and shape control of soft robots is proposed.The main research elements of this paper are as follows:A vision-based approach to soft robot shape and posture measurement is proposed,as the complex multi-morphological variations of soft robots make it difficult to measure their kinematic parameters.By detecting feature points that characterise the shape of the soft robot,the shape of the robot is reconstructed and a kinematic model is built to find the geometric mapping between deformation and kinematic parameters to measure bending angleθ,deflection angleφ,elongationΔL and position coordinates（x,y,z）.A vision measurement platform was built to perform camera calibration,shape reconstruction and motion parameter measurement experiments.The experimental results show that the root mean square error of the end position of the soft robot is less than 6 mm,the position deviation of the overall shape is 1.67 mm,and the average error of the bending angle and deflection angle are 3.6°and 3.4°respectively,indicating that the vision measurement method can accurately estimate the shape and motion parameters of the soft robot.An inverse kinematic model control method considering the robot under the coupled perturbation of gravity and motion is proposed for a two-segment pneumatic soft robot as the control object.According to the structure and motion characteristics of the soft robot,the forward and inverse kinematic models of the two-segment soft robot are established using the constant curvature method,and the non-linear mapping functions between the input air pressure of each channel in the drive space and the output position in the task space are obtained.A two-segment soft robot position control platform based on visual feedback is built,and the point-to-point position control of the two-segment soft robot is carried out using stereo visual feedback of the end position of the soft robot.The experimental results show that the control error based on the inverse kinematic model is less than 3.5 mm,and can accurately move to the desired position with less visual feedback,and the angular error and position error are less than 1°and2 mm respectively,indicating the effectiveness and feasibility of the control strategy.In order to achieve the movement of a two-segment soft robot to a desired form,a vision-based method for controlling the form of a two-segment soft robot is proposed.To achieve accurate control of its morphology,shape control is changed to control of its feature points,a constant curvature method is used to characterise the feature points,its shape feature parameters are applied to describe the soft robot’s 3D shape profile,a shape control algorithm based on visual feedback of the constant curvature feature parameters is designed,and a mapping model between drive and shape feature parameters is established.In order to illustrate the effectiveness of the control method,experiments on the L-type and S-type shape control of the desire are conducted.The experimental results show that the endpoint position error and angle error are less than 2 mm and 1°respectively,and the curvature error is less than 0.0001 mm^-1,indicating that the 3D shape control algorithm has a good effect on tracking the desired 3D shapes of L-type and S-type shapes.