A fuzzy self-organizing map (SOM) network isproposed in this paper for visual motor control of a 7 degreesof freedom (DOF) robot manipulator. The inverse kinematicmap from the image plane to joint angle space of a redundantmanipulator is highly nonlinear and ill-posed in the sensethat a typical end-effector position is associated with severaljoint angle vectors. In the proposed approach, the robotworkspace in image plane is discretized into a number offuzzy regions whose center locations and fuzzy membershipvalues are determined using a Fuzzy C-Mean (FCM)clustering algorithm. SOM network then learns the inversekinematics by on-line by associating a local linear map foreach cluster. A novel learning algorithm has been proposedto make the robot manipulator to reach a target position. Anyarbitrary level of accuracy can be achieved with a numberof fine movements of the manipulator tip. These fine movementsdepend on the error between the target position and thecurrent manipulator position. In particular, the fuzzy modelis found to be better as compared to Kohonen self-organizingmap (KSOM) based learning scheme proposed for visualmotor control. Like existing KSOM learning schemes, theproposed scheme leads to a unique inverse kinematic solutioneven for a redundant manipulator. The proposed algorithmshave been successfully implemented in real-time on a7 DOF PowerCube robot manipulator, and results are foundto concur with the theoretical findings.
- Redundant manipulator
- Inverse kinematics
- Visual motor control
- Fuzzy self-organizing map network