Broad area/overview

The broad area of this paper is about reducing trajectory tracking errors for MIT's Serial Link Direct Drive Arm robot. This group studied a range of controllers which included analog PD control applied independently at the joints to feedforward and computed torque methods which incorporated full dynamics.

Specific Problem

• This paper seeks to address the problem of having accurate modeling and prediction of the robots joints positions and torques. In 1989, there were not any viable commercial options which offered what the controllers proposed at the time required: joint torque.

Solution Ideas

• This group performed a set of 2 experiments with different control strategies. The first set was a hybrid control approach where each joint has an independent analog servo with the position, velocity, and feedforward commands generated by a microprocessor. The second set was using digital servos and one Motorola 68000 based microprocessor. They sought to compare the online compute torque approach to the standard PD as well as the feedforward approach with digital servos.

• The group did not have the powerful CAD software that we have available today. They had to do much more work to generate the inertial values of the robot links. They used a least-squares algorithm to estimate the inertia.

• In their results they present trajectory tracking error plots for their experiments with both controllers. Their results showed that with either model, dynamic compensation improved trajectory accuracy significantly. They found that giving the robot a reference velocity for its computation greatly increased the accuracy of the results as well.

• This group ultimately in the end seemed limited by the equipment they had available to them at the time. They reported that though their results were good, they had problems which limited their results. Inconsistencies in the position and velocity sensor, torque estimation errors, and too low a sampling rate were what the researchers felt hindered them. These all seem like hardware limitations with the IMUs, motors, and microprocessor. The sophistication of their hardware was not enough such that their models and commands were as strictly enforced as needed.

Comments

• This paper is actually working on something very similar to what we were modeling in Julia with our first assignment.

• These researchers are doing some important early work on trajectory tracking and error and unfortunately were very limited by their available hardware and electronics

• This paper really shows how much this field is exploding and will continue to grow. There is so much work to be done that seems to have been planned and desired for a very long time and the technology to finally support those ideas and control strategies is coming online and becoming more widely available.

Recent Papers

• Passive Internal Model Based Repetitive Control of Robot Manipulators