Energy was the plan, time became the metric
Project 2 of our robotics module at UCD was a practical study in a group of five. The task was to program a pick-and-place routine on a KUKA robot, set a baseline, change one thing at a time and see what it did to energy use. I designed the path, programmed it on the pendant and ran the optimisation.
Energy needed a data acquisition box and clip-on sensors, and that setup was not ready on the day. We were told to compare runs without it. So the question became simpler: how much shorter can a cycle get if the robot does not move any faster? A shorter cycle does not prove lower energy, but a robot draws power the whole time it runs, so the two usually move together. I did not measure that and I will not claim it.
This sits next to my work on robots in the electronics industry and the KUKA.Sim cell. There the robot was virtual. Here it was a real arm with a real pendant.
A real arm, four bottles
The robot was a KUKA KR10, a six-axis industrial arm on a KR C4 controller. I wrote the program in KRL on the smartPAD teach pendant and ran it in T2 mode, where the robot moves at programmed speed only while the Start key is held.
The task was short. Start from home, open the gripper, pick four bottles one after another, place each at a drop position, return home. I also designed the bottle holder in Autodesk Fusion 360 and supplied the CAD for it, so the bottles sat in the same place on every run.
| Item | Detail |
|---|---|
| Robot | KUKA KR10, 6 axes |
| Controller | KR C4, KUKA System Software |
| Language | KRL, written on the smartPAD |
| Mode | T2, test mode at programmed speed |
| Task | Pick 4 bottles, place each at a drop point, return home |
| Programs | One baseline, one optimised |
| Lab | LIMAR lab, UCD |
Get it working, then look at it
The first program only had to work. I ran it at 75% and at 100% of the pendant's speed override, which is the speed setting on the pendant. Then I watched it and wrote down what looked wasteful.
- Approach and retract points sat higher and further from the bottles than they needed to.
- The arm almost stopped between points instead of flowing through them.
- One axis kept rotating back and forth for no reason.
- The trip from pick to drop was longer than it had to be.
Better path, same speed
Speed and acceleration stayed exactly where the 100% baseline had them. Any gain had to come from the path.
Closer approach and retract points
Smoother via points
No wasted axis rotation
Shorter pick-to-drop path
Two clocks, one answer
Cycle time here means the robot leaving home to the robot returning home, covering all four bottles. I timed it twice. A phone stopwatch gave a quick number, good to about 0.2 seconds. The lab video, timed frame by frame at 30 frames per second, gave a number anyone can check, good to about 0.03 seconds.
The two methods agreed within 0.4 seconds. The stopwatch read about 0.3 seconds short each time, which is the delay of pressing a button by hand. The video was not filmed for the 75% baseline, so that row has a stopwatch time only.
About two seconds, from the path alone
| Run | Stopwatch | Video, frame by frame |
|---|---|---|
| Baseline, 75% speed | 69.99 s | not filmed |
| Baseline, 100% speed | 66.88 s | 67.17 s |
| Optimised, 100% speed | 64.76 s | 65.13 s |
Going from 75% to 100% speed saved 3.1 seconds. The path changes saved 2.1 seconds with no speed increase at all, about two thirds of what a 25 point speed rise gave. Against the 75% baseline the optimised run at 100% was about 7.5% faster in total.
A speed rise helps less than you would expect. The gripper action and the settling at each point do not get faster when the override goes up, so part of every cycle stays fixed.
What this changed in how I work
Path design is a lever people skip. Fewer stops, less vertical travel and no wasted joint motion cut time, and they also cut wear on the arm. If this cell ran around the clock, two seconds a cycle would turn into a lot of extra parts per shift for free. That is lean thinking applied to a robot: remove the motion that adds nothing.
Measure with two methods. A hand stopwatch is fine when something else checks it. And plan for the lab not behaving. The energy rig was not ready and the controller threw a fieldbus warning (KSS10094, a communication fault with a device on the bus). The warning did not stop the robot, but it reminded me that real equipment rarely follows the plan, and a sound backup metric kept the study useful.