|Ph.D Student||Valency Tomer|
|Subject||Biologically Inspired Motion Pattern Generator for Robotic|
Task Planning and Control
|Department||Department of Mechanical Engineering||Supervisor||Professor Miriam Zacksenhouse|
Biological systems present important advantages over robotic systems in movement planning and execution. These advantages are manifested in the ability to adapt motion to geometrical and dynamical uncertainties of the environment, and to synchronize the coordination of multiple robots. The advantages of the biological systems over robotic systems do not result from any superior physical qualities. On the contrary, in most of the physical parameters, including speed, repeatability and accuracy, robotic systems outperform the biological systems. Rather, the advantages of biological systems seem to stem mainly from the employed strategies for trajectory planning and adaptation.
The work reported here developed a biologically inspired method for planning and executing robotic tasks termed Online Path Generator (OPG). The method generates an online trajectory command in the robot joint space to be followed by the robot’s internal controller. The OPG includes two parts: a nominal trajectory generator and a trajectory adaptor. The nominal trajectory generator is based on Central Pattern Generator (CPG) emulation, which generates a nominal trajectory in the robot joint space that is characterized by zero phase difference between all joint movements. The trajectory adaptor modifies the nominal trajectory by altering the joint phase differences and the overall phase in response to task-space feedback. The feedback law minimizes an error function that includes both the deviation from the nominal trajectory and deviation from the task goal, as reflected by the feedback. The use of the phase differences and the overall phase as a control objective is, as far as we know, unique in robotics.
By online adaptation of the desired trajectory to the actual measured environment, the OPG facilitates the performance of tasks that involve significant uncertainties in the environment. Furthermore, when coordinating the operation of more than one robot the OPG coordinates the motion of the different robots via feedback and eliminates the need to connect all the robots to a common controller (as in a centralized control system).
In redundant systems, the OPG method employs a form of “weighted generalized inverse” to transform the instantaneous trajectory command from the task to the joint space. The OPG method uses a position controller in the joint space to track the desired instantaneous trajectory as in “Resolved motion rate control” method. By using a nominal default trajectory the OPG method simplifies the planning of all robotic tasks and facilitates the planning of complicated robotic tasks.
The applicability of the OPG method is demonstrated in this work by simulations and experiment.