Design and Development of a Redundant Manipulator Arm

Members:

• Mitul Saha
• Sayeed Sanaullah
• Siddharth Kumar

Supervisors:

• Dr. Bhaskar Dasgupta
• Dr. Susmit Sen

Aims and Objectives:

Plants which require maintenance and inspection of sites which are inhospitable to human beings due to number of reasons like, high temperature, presence of toxic or radioactive substances, demand such manipulators. Redundancy is required for Obstacle Avoidance (A manipulator is said to have redundant links, if it has degrees of freedom more than required to reach a point. In the case of parallel plane and serial manipulators, a manipulator is said to have redundant links if it has more than 2 degrees of freedom).

The situation is complicated as many of the inspection sites are difficult to access manually (as the workspace may be very congested) as well as by mobile robots (as the terrain may not be even). In such a scenario, the inspection sites can be accessed in a serpentine fashion.

The present project is intended to meet such demands of a complicated environment, by harnessing the potential of redundant manipulators to move in a serpentine fashion.

Overview:

A robot manipulator having more degrees of freedom than essential for accessing a point in space is called a redundant manipulator. Such redundancy helps in obstacle avoidance, singularity avoidance, workspace enhancement, performance optimization and improvement in reliability leading to safer operational situations. The redundancy will be utilized for reaching and dexterous maneuvering at locations that are difficult to reach, but nevertheless require periodic maintenance. Path Planning and Obstacle Avoidance strategies are required. Subsequently various control schemes will be designed for the system and comparatively evaluated in order to finalize the most suitable strategy.This entire exercise is expected to lead to a systematic methodology to design a redundant manipulator for any workcell that may involve inaccessible locations.

Steps Involved:

• Path Planning and Obstacle Avoidance:

  Intermediate pathpoints for the end effector is required for moving from start to goal point; also the manipulator should avoid obstacles. Initially we tried to test the traditional "Potential Field approach" graphically in OpenGL. It is very difficult to find a potential function applicable to all obstacles. We concluded that potential field approach is not at all practical, as the manipulator tends to vibrate near the obstacles.
  
  Next we have explored through another approach called: "Visibility Graph Search". This along with Inverse Kinematics has proved to be quite feasible. Though Inverse Kinematics for redundant manipulators is not easy to handle, one can simplify the Inverse Kinematics for 4 DOF parallel plane, serial manipulators.
  
  Finally we will also demonstrate Path Planning and Obstacle Avoidance using Genetic Algorithm. This is an original contribution from our side towards Path Planning and Obstacle Avoidance. We have found by the use of Genetic Algorithm one can easily eliminate the problems of Potential Field Approach and Inverse Kinematics and the manipulator can very easily navigate through very narrow regions without getting stuck.

• Manipulator/Workcell Design and Development:

  In our four-link manipulator, links would be made of a light and sturdy composite material called "perspex". The links would be driven by servomotors. The following figure gives the overall look of our proposed design. Roller supports to links are given to avoid sagging. The design is portable and can be easily moved from one place to another.

  

  To start with we try our manipulator in the workcell shown in first figure. But our manipulator should work for any workcell. Every time a new workcell is given,we would only change the description of obstacles in our Path Planning program, rest will remain same.

• Control:

  Keeping the obstacles under consideration, redundancy resolution schemes will be developed to maximize the dexterous performance of the manipulator. We need to find the time behavior of the force and torques to be delivered by the joint actuators so as to ensure the execution of the desired trajectories. Control of end-effector motion demands an accurate analysis of the characteristic of the mechanical structure, actuators etc. Therefore manipulator control is ensured to the closure of feedback loops, by computing the deviation between the reference inputs and the data provided by the proprioceptive sensors. A feedback control system is capable to satisfy accuracy requirements on the execution of the prescribed trajectories. Servo and model-based control strategies will be developed and experimentally evaluated with respect to their performance and cost in order to finalize the most suitable control scheme. In fact much of the challenge of the design and its efficient use focuses on this aspect.

  The data generated by the path-planning program is fed to control software called LABVIEW. This software converts the digital signal to analog using a control card called FLEX MOTION PC1-6c. The specifications of this control card are shown after few pages. The signal is fed to motor, which drives the manipulator using the gear head. The encoder senses the position of the joints at regular interval of time. This provides data for feedback control in terms of velocity and position. It is converted to analog signal using the same control card. Thus the feed back loop is completed and the desired motion is obtained.

• Calibration and Testing:

  The project does not end with Controller design. Kinematic Calibration is also required at the end, after the manipulator has been developed along with the Path Planner and the controller.

Current Status:

Till now we have completed the theoretical aspects, like Path Planning and Mechanical Designing. However graphical testing of the Path Planner is still going on. Attempt to make it as general as possible, i.e. independent of workspace, is going on. Manufacturing of the components has started. This project is expected to complete by February end. If successfully completed we will go for its publication. Meanwhile the page will be updated from time to time till the end of the project.