From February 1987 to June 1995, I worked in the Robot & Automation Research Institute in Beijing, China. My job title was "Engineer Assistant" (February 1987 to February 1988), "Engineer" (February 1988 to August 1993), and "Senior Engineer" (August 1993 to July 1995), and my major work was mechanical system/parts design, though I also dealt with system control issues from time to time. The following are some important projects I involved.
The HRGP-1A model was a 6-DOF (Degrees Of Freedom) electro-hydraulic servo-valve controlled industry robot designed for painting task in automobile manufacture industry. Its mechanical structure was similar to the DeVILBISS TR-4500 robot: with three big linear cylinders actuating the base, vertical arm and horizontal arm respectively, and two small linear cylinders plus one rotary actuator driving the tri-jointed flexible wrist. As the components' gravity was well compensated to a balanced status with a set of spring mechanisms, the whole arm structure could be easily guided into any pose.
For the robot control system, each of the six controlling system sets had a servo-valve as its actuator's adjusting element and a resolver as the actuator's piston position sensor. The basic control strategy was digital-PID controlling method handled by an EISA industrial computer (80386) system. The computer system had a master-computer to complete overall system management (data processing or noise-filtering, path interpolating, PID controlling, parts identification and system communication etc.) and a slave-computer processor MCS-8751 to fulfill the 6 channel servo-system's I/O function.
In the model HRGP-1A robot research and development, I was mainly involved in the robot arm parts' mechanical design and then the manufacturing services. I did part of the system programming with assembly language. My programming design and maintenance work included processing the data with a Butterworth filter and adjusting PID parameters through experiments.
My paper "Error Analysis of HRGP-1A Painting Robot" (in Chinese language) addressed the needs of HRGP-1A Robot research and development. The following is the English abstract of the paper:
After all possible sources of the HRGP-1A Robot pose errors were examined, the main error sources, i.e., the arm driving errors, were focused in this paper. First, the kinematical equations were derived to relate the robot pose to its joint angles. Then the transformation matrixes, which described the robot pose of end-effectors versus cylinder actuating piston position, were derived. Finally, by using the concepts of differential translation and rotation, the HRGP-1A Robot's pose control error model was successfully built. The analysis method and results were used as a guideline for our robot control parameter adjustment and for further system improvement.
The DSS project built a prototype of specially designed seat to simulate the movement of a vehicle according to the plot of being played movies in order to entertain the audience sitting in it. We built two prototypes, one was 3-DOF (Degree Of Freedom) and the other was 6-DOF. They were both electro-hydraulic servo-valve controlled systems, with servo-valves and linear cylinders as their actuators and potentiometers as their position feedback components. The basic control method was digital-PID, implemented via a computer system using MCS-8098 as its CPU.
My task was: (1) to analyze the mechanical structure in order to derive the kinematical relationship between the seat pose (position and orientation) and the hydraulic cylinder pistons' position; (2) with this relational equations, to analyzed the movie/video plots and design movement sequences, and to generate desired cylinder's piston position-versus-time series data.
My report "Motion Analyzing and Calculating of DSS" (in Chinese language) was about the inverse kinematical analysis of the system. The following is an English summary of it:
The research project of DSS (Dynamic Simulation Seat) was to build a system that drove the specially designed seat to follow the designed movements according to the being played movie plots, for the purposes of entertaining the audience sitting in it. The system that we were studying had 6 degrees of freedom.
In order to drive the seat into the specific pose, the relationship between the seat pose and the driving cylinder's piston positions had to be known first. This report dealt with the so-called "inverse kinematics" problems in robotics. However, the DSS upper-part driving configuration differentiated it from ordinary robot structures whose kinematical equations can be readily derived with the classical Denavit-Hartenberg method: the seat was bolstered by three driving cylinders, whose piston positions jointly determined the pitch and roll orientation and up-and-down moving position of the seat.
In this report, the seat's driving mechanical structure was analyzed and the necessary reference coordinate systems were selected first. Euler coordinate transformation method was used in the equation derivation, and the motion compensation and constraints were studied in detail. The equations of seat position and orientation versus driving piston's displacements were derived and the computer calculation procedures were given out. This work served the DSS system programming.
A robot with vision sensors was designed in our institute to fulfill the task of filling the swinging rocket fuel tank in launch tower . It was a 3-DOF (Degree Of Freedom) robot with a Cartesian-coordinate structure (YZX kinematical chain). The vision tracing system was made up of two sets of single-array CCD as its sensors (for Y and Z direction respectively). By the time I left, the experiment and pre-design were completed.
I was responsible for the system design and experiments. My major work was mainly in the vision system software design and the system control. The following is an English summary of my report (in Chinese language):
When all the on-ground-testing have been finished, the satellites-carrying rocket is erected in a launch tower. To fill the swinging rocket fuel tank in the tower is a dangerous and difficult task because the fuel is usually very poisonous and extremely cold. We intend to design a robot-fueling system to fulfill this task.
According to the launching service people, the swinging of the rocket in the launch tower is mainly in an elliptical patterns, and a 3-DOF (Degrees-Of-Freedom) Cartesian-structure hydraulic-powered robot with vision detecting system is suitable for such a task. The system pre-design is completed, and this report is about the project feasibility analysis.
To test the whole system design, a simplified system with 1-DOF equipment set has been set up in our laboratory. It contains two parts: the simulator and the tracer; both are driven by a hydraulic pump. Apparatus BT6 is adopted to generate oscillating signals and to test system response characteristics.
The rocket motion simulation loop consists of a potentiometer, a V/A converter, a servo-valve and a linear cylinder actuator set. The V/A converter receives oscillating signals and the actuator drives a small vehicle, which bears the target object moving forwards and backwards in simulating one direction motion of rocket swinging.
The robot vision tracking loop also consists of a V/A converter, a servo-valve and a linear cylinder actuator, but its position feedback unit is switch-able. When the potentiometer is used as the position-feedback element, this close-loop system can control its actuator to drive another small vehicle, which bears the photoelectric sensor or the CCD camera, to a given position. When the photoelectric sensor or the CCD camera detecting unit is used as the close-loop position-feedback control element, this closed loop just resembles the designed robot vision tracing system in 1-DOF direction.
With all the equipment, the designed Robot Vision Tracing was experimentally proved feasible. This report covers the detailed description of principle, procedure and our experiment results. Also covered in the report is the system dynamic characteristics analysis and system calibration study.
This report can be used as a guide in prototype robot tracking system design.
The HRGB-1 Robot was designed to meet the needs of rocket manufacture process. The prototype was finished and delivered to the CALT (China Academy of Launch Vehicle Technology) to take the place of workers to assemble the filling-stuffs, which are too heavy and difficult to be manually handled under the conditions of very limited space. The robot has a Cylinder-coordinate structure (CZR kinematical chain). In this engineering project, I undertook the mechanical design of the Vertical Arm Assembly and took a part in designing of electronic control system.
In the Robot & Automation Institute, I also participated the research and development of other commercial products. My major work was mechanical subsystems' or parts' design and their manufacturing process technical services.