Research On Multi-Dof Flexible Manipulator For Shallow Lunar Regolith Sampling

Lunar exploration has always been a hotspot and it is of great importance in space exploration. The sampling technology is critical in the process of exploration. In our country, the "Chang'E" project is on its way and the sampling and analyzing task is included in the project. However, the strict restraints coming from the severe circumstances on the lunar surface requires the sampling equipment has small volume, light weight and small power consumption. In the thesis, a novel multi-DOF flexible manipulator is proposed for shallow lunar regolith sampling. By combining the rigid link and the flexible link, the sampling manipulator has some significant merits, such as small shrinking volume, big working space, and small power consumption. The corresponding electronic control system is made to detect and control the manipulator. Also, the kinematic analysis, the dynamic control, and the in-situ vibration signal identification are carried out for the special mechanism. And the error modeling is done for the developed manipulator.Firstly, based on the original one-DOF flexible sampler, a novel rigid-flexible link combined manipulator is designed for the shallow lunar regolith sampling task, which has three more DOFs. Thus, the merits of the original sampler is maintained and moreover, the working space is enlarged and the sampling spot is selective in the sampling area. The DSP TMS320F28334 and LabWindows/CVI software platform based detecting/controlling circuit is designed and manufactured which has the function of detecting the motor current signals, controlling motors in each joint, generating the vibration control signals, controlling the vibration motors, acquiring the motor current signals, and controlling the sampling spoon, as well as acquiring the vibration signals of the sampling head and sending them to the upper computer for further analysis.For the rigid-flexible link combined manipulator, the kinematic analysis and the flexible displacement compensation are done to improve the accuracy of kinematic control. Based on the kinematics, the dynamic model is built after reducing the DOF of the manipulator to two according to practical requirements. The flexibility of the manipulator is considered and the modal analysis for the flexible link is performed in the process of building the dynamic model. The feasibility of the dynamic model is verified. Based on the dynamic model, the linear quadratic control method is introduced and modified to control the vibration of the flexible link. Lastly the theoretical curve and the practical curve of the angular displacement of the 2nd joint are obtained and compared, which verifies the effectiveness of the control method.In order to highlight the merits of the flexible manipulator, i.e., small shrinking volume, the mechanism is improved based on the multi-DOF flexible-rigid link combined manipulator. The rigid link is replaced by a 2 DOFs base so as to further shrink the volume. The round holes are punched in the middle line of the flexible link to match the driving gear. Then the flexible arm-driving mechanism increases the drilling force of the sampling head and improves the reliability of the manipulator. The error modeling is performed for this improved mechanism, which includes manufacturing errors, input variable errors and joint clearance errors. Lastly, the influence on the location errors of the sampling head from each error source is analyzed in detail.Using the vibration method to improve the effectiveness of sampling becomes feasible under the condition that the sampling arm is flexible, so the influence coming from the vibration is reduced to minimum. Moreover, the function of the flexible sampling manipulator can be largely extended by acquiring the vibration signals and analyzing them. The sampling head-rock contact identification and the regolith bulk density measurement can be done by several steps of vibration signal processing, i.e., the signal trend removing, wavelet filtering, Periodogram/Welch feature extraction, and lastly, the support vector machine/K nearest neighbors and General Regressive Network identification. And the accuracy for these two kinds of identification are 94.5% and 89.22%, respectively.At the end of the thesis, a series of experiments was carried out to verify the feasibility and the performance of the manipulator. And the results proved the effectiveness of the control method.