Introduction
The goal of this work is to extend the payload capability of a Puma 762 robot beyond its established 45-lb (20-kg) limit to its maximum potential.
The optimal control problem is formulated as maximizing the lifted payload subject to physical constraints and is then solved by gradient methods starting from an initial feasible path.
To demonstrate our results we try to find the largest weight that the Puma could lift from the downward posture to the upward posture.
Failed Lift
Next we attempt to have it move 90 lbs on the same path.
The attempt fails, demonstrating the 45-lb limit.
1 DOF
We then use the optimal algorithm to find the path that maximizes the payload.
In order to check our solution, weights are added until the arm could no longer lift them on a specified trajectory.
Note how, in this case, energy is pumped into the system during the swinging motion.
This solution lifts a maximum of 70 lbs.
3 DOF
When more DOFs are added to assist in the motion, the individual joint torques are reduced.
In this case we allow all 3 wrist joints to move.
The Puma wrist goes through a singularity.
While moving through the singularity, some joint torques are reacted by the structure and not the actuators.
6 DOF:
Sol. 1: 115 lbs, Sol. 2: 125 lbs, and Sol. 3: 145 lbs
Finally, we allow the algorithm to vary all 6 joints.
We form several local minima to the nonlinear optimal control problem.
Three different solutions are shown.
The last solution increases the payload capacity of the Puma by more than a factor of three.
Note that, again, the paths move through sigularities.
References:
''Weightlifting Motion Planning for a Puma 762 Robot,''
C,-Y. E. Wang, W. K. Timoszyk, and J. E. Bobrow,
IEEE International Conference on Robotics and Automation, Michigan, May 1999.