Developing a LabVIEW-Controlled Hyper-Redundant Manipulator Arm for a Cluttered Environment

Author(s):
Himanshu Mishra - Center for Robotics, Indian Institute of Technology
Susmit Sen - Center for Robotics, Indian Institute of Technology

Industry:
Electromechanics/ Electrotechnics, Machines/Mechanics, Imaging Equipment

Products:
Motion Control, LabVIEW

The Challenge:
Planning a path and control of a hyper-redundant robotic arm, consisting of simultaneously controlled servomotors and stepper motors to operator in a cluttered environment.

The Solution:
Using LabVIEW and FlexMotion software to control the robotic arm.

Every robotic arm is specified by its degree of freedom (DOF) it has. DOF is the independent coordinate plane or orientation that a joint or end effector can move. The number of independent variables determines DOF in the system. End effector is the last link, which is usually a tool, or a gripper used to perform work on an object. A robotic arm is redundant if it has more degrees of freedom than required to reach a point. In the case of serial manipulators, a manipulator is redundant if it has more than two DOF.

There is a strong demand for robotic manipulators in plants that require maintenance and inspection of sites that pose threats to humans because of high temperature or presence of toxic or radioactive substances. We require redundancy for obstacle collision avoidance. The situation is complex because many of the inspection sites are difficult to access manually because of congested workspace or by mobile robots because of uneven terrain. In such a scenario, we can access the inspection sites in a serpentine fashion. With the present project, we met the demands of a complex environment by harnessing the potential of redundant manipulators to move in a serpentine fashion.

We needed to design a robotic arm, so it could reach any position to avoid obstacles with high flexibility. The hyper-redundant manipulator used a large degree of redundancy to improve the manipulator performance in a complex and cluttered workspace. However, because of the high degrees of freedom, path planning, synthesis and control were difficult problems to tackle.

Path Planning
We needed intermediate path points for the end effector to move from start to finish. The manipulator also needed to avoid obstacles. Initially, we tried to test the traditional “potential field approach” graphically in OpenGL. The potential field approach defines a potential function over the free space. Ideally, it repulses the robot away from the obstacles and attracts it towards the goal. It is difficult to find a potential function applicable to all obstacles. We concluded that potential field approach was not practical because the manipulator vibrated near the obstacles. Next we explored another approach – the “visibility graph search.” While the visibility graph method is the roadmap method that constructs the shortest path, it is a path, which is collision-free. When a robot travels on the roadmap method, it stands the risk of encountering obstacles along the way. This method, along with the inverse kinematics approach, was quite feasible. Though it was not easy to handle inverse kinematics for redundant manipulators, we simplified the inverse kinematics for four DOF planar serial manipulators. We discarded the visibility graph search method because of the high computational time it required. Finally, we considered the single-query bi-directional probabilistic roadmap method (PRM). The probabilistic roadmap, which is a network of simple paths (usually straight line segments in configuration space) generated by picking free configurations at random. The PRM provided us a number of path solutions, but we selected the optimal path based on minimizing the number of links and maximizing the quality of reach among the paths that are able to reach the goal. In the end, we obtained a table of angles that displayed the movements of each joint to reach the target position. This table is called the theta file.

Manipulator and Obstacle Design and Development
In our four-link manipulator, links consisted of a light and sturdy composite material – “perspex.” Servomotors and a stepper motor drove the links. The manipulator had four links with a planar design.

Initially, we tested our manipulator in a simple obstacle environment even though our manipulator was designed to work in any reasonably complicated environment. Each time a new obstacle was introduced in the path of the arm, we only changed the description of obstacles in our path planning program, which includes the number of vertices of the obstacle and the coordinates of the vertices) and the rest remained the same.

Controlling the Hyper-Redundant Arm
Using LabVIEW and the NI high performance motion controller, we controlled our hyper-redundant arm. We needed to find the time behavior of the force and torques to be delivered by the joint actuators to ensure the execution of the desired trajectories. Control of end-effector motion demanded an accurate analysis of the characteristic of the mechanical structure and actuators. This ensures manipulator control for the closure of feedback loops by computing the deviation between the references inputs and the data provided by the proprioceptive sensors. A feedback control system can satisfy accuracy requirements on the execution of the prescribed trajectories. We developed and experimentally evaluated servo and model-based control strategies with respect to their performance and cost to finalize the most suitable control scheme. In fact, much of the challenge of the design and its efficient use focused on this aspect.

The theta file, generated by the path-planning program, was fed to LabVIEW control software. LabVIEW converted the digital signal to analog using the NI high performance motion controller card. We fed the signal to the motor, which drove the manipulator using the gear head. The encoder sensed the position of the joints at regular intervals of time. The encoder is a sensor that converts angular or linear position data into digital form. Optical-based encoders are most common. This provided data for feedback control in terms of velocity and position. The same control card converted data to analog signals. Thus, we completed the feedback loop and obtained the desired motion.

We used three maxon DC brush motors with built-in quadrature encoders and one controlled-loop stepper motor. We prepared a circuit on a general purpose PCB using Op-Amp 12A to interface servomotors and encoders. We made a separate driver for the single stepper motor, which worked open loop. The servomotors had a rating of 12288 counters per rotation, which provided us with the accuracy of 0.029 degrees per count.

We found the autotuning feature in the Measurement & Automation Explorer effective in decreasing the system response time and reducing the steady-state error. This autotune feature saved the time spent on finding the PID values for fine-tuning the system. This feature is rare in custom controllers. We made the control panel of hyper redundant manipulator extremely user friendly for a semiskilled operator by exploiting the rich control palette library in LabVIEW. The control panel had a manual/automatic select button that we used to make the arm move by entering the joint angle values manually or making the program read a theta file, which the PRM code generated. We moved the arm manually by entering the joint angle values for the purpose of testing the accuracy of the arm from time to time. With the “back to initial position” button, we made the arm return to initial position, following the same theta file but reading it from the bottom with a reverse direction. The control panel also included a chart display of the steps and counts in relation to the time of stepper and servomotor used in the system.

We initialized two vector spaces in the block diagram because we needed all motors to move at the same speed and acceleration. As a primary requirement of path planning, we needed our program to read one row of the theta file at the same instant and perform as a single move. Because of this requirement, we kept the blend factor zero, so one move finished before the other move started. We changed the velocity and acceleration of motors by making changes in the block diagram.

Scope for Future Work
The present manipulator has four-degrees of freedom, consisting of three servomotors and one stepper motor actuator. Considering the limitation of a single NI high performance motion controller card, we can always increase the DOF to six DOF. If we wish to increase the DOF further, we can add another NI high performance motion controller card to the existing system. The present path planning code has a computational time, which varies with the complexity of the obstacle. Since PRM is a stochastic search method, the time for finding the path exponentially increases with the increase in the robot’s DOF. The motion planner is useful for manipulators with up to 15 degrees of freedom.

We can easily integrate a path-planning code optimized for computational time into a LabVIEW block diagram using code interface note (CIN) and use this code for on-the-spot programming. Using the NI-IMAQ vision card, we can input the coordinates of the obstacle through a camera mounted vertically on the workspace without manually entering the coordinates of the obstacle each time we run the path planning code. This will add a vision card to the existing system, which already has a motion-control card in it. The basic idea of using a camera mounted on top of the workspace is to convert the existing system to work in a mobile obstacle environment.

Reducing Development Time
We tested LabVIEW and the NI high performance motion controller card for a four DOF hyper-redundant manipulator, and they provided the flexibility of graphical programming and reduced development time. The integrated motion-control capability was critical to reaching our goal. Four degrees of freedom is an acceptable start for testing a hyper-redundant manipulator, and increasing the degrees of freedom is a welcomed issue. LabVIEW would be an obvious choice for the control of hyper-redundant manipulators in the future.

For more information, contact:
Susmit Sen
E-Mail: sens@iitk.ac.in
Himanshu Mishra
E-Mail: hmishra@iitk.ac.in
Center for Robotics
Indian Institute of Technology Kanpur, India

Browse All Case Studies »