Home About Expertise Projects Blogs Contact
RoboticsKUKA KR10Cycle time

Faster without
speeding up

A KUKA KR10 did the same pick-and-place job 2.1 seconds faster. Nobody turned the speed up.

CN
Chandan Naveen
·November 19, 2026·6 min read·Robotics · Automation · Manufacturing
KUKA KR10 six-axis industrial robot arm used for the pick-and-place cycle time project at UCD

The robot did the same job 2.1 seconds faster and nobody turned the speed up. Going from 75% to 100% speed saved 3 seconds. Changing the path saved 2.1 seconds with no speed change at all.

The project

This was Project 2 in MEEN41350 Introduction to Robotics, part of my MEngSc in Robotics and Intelligent Manufacturing at University College Dublin. It is a practical study worth 20%, done by Group 11 (five of us) in the LIMAR robotics lab in November 2026. The brief was energy optimisation of an industrial robot: build a pick-and-place task, set a baseline, change parameters such as layout, approach and retract offsets, payload data, motion blending, speed and acceleration, and measure the effect.

The robot was a KUKA KR10, a six-axis industrial arm, programmed in KRL on the smartPAD teach pendant. The task: start at home, open the gripper, pick and place bottles 1 to 4 one by one, return home. GROUP112026 was the baseline program and GROUP112026OPT the optimised one. We ran in T2 mode, which runs at programmed speed and moves the robot only while you hold Start. Speed came from the program override, a pendant setting that scales every move as a percentage of programmed speed.

My part was designing the robot path, programming it on the pendant and doing the optimisation. I also designed the bottle holder in Autodesk Fusion 360.

The bottle holder

A taught robot point has no vision and no correction. If a bottle sits 5 mm off, the gripper still goes to the taught spot. So the parts must be in the same place every cycle. The holder fixes all four bottles in set positions. It measures 190 x 60 x 33 mm, with a 3 mm base plate, 8 mounting holes and four cups 30 mm tall, spaced 42.5 mm apart. The STEP and STL files are on GitHub.

The plan changed

We meant to measure energy with an NI USB DAQ box and clip-on current and voltage sensors. The setup wasn't ready, so we compared cycle time instead, which the brief allowed. The controller also showed a fieldbus warning (KSS10094). It didn't stop the robot, but it was a good reminder that labs rarely match the plan. Pick a backup measure you can trust.

What was wrong with the baseline

The baseline wasted motion in four places. Approach and retract points sat higher and further from the bottles than they needed to, so every pick had extra lifting. There were near-stops between points. One axis kept rotating back and forth for no reason. And the pick-to-drop path was longer than it had to be.

What I changed

I moved the approach and retract points as close to the bottles as was safe. I re-placed the via points by hand for smooth motion, without using KUKA's blending setting. I removed the wasted axis rotation, shortened the pick-to-drop path and smoothed the return to the next pick.

Speed settings stayed the same as the baseline, so only the path changed. One change, one measured effect.

How we measured

Cycle time from home to home, all four bottles, by stopwatch. That is good to about plus or minus 0.2 s. The 2.1 s saving is roughly ten times that error, so the result holds.

RunSpeedCycle timevs baseline 100%
Baseline50%130.33 s+94.9%
Baseline75%69.99 s+4.7%
Baseline100%66.88 sreference
Optimised100%64.76 s-3.2%

What the numbers say

3.2%
Cycle time saved by path changes alone
2.12 s
Saved per cycle at the same 100% speed
3.1 s
Saved by going from 75% to 100% speed

Path changes alone saved 2.12 s, about two-thirds of what the 75% to 100% speed jump gave, and the robot ran no faster.

Speed gains flatten out. Going from 50% to 75% saved about 60 s. Going from 75% to 100% saved 3 s. Gripper action and settling at each point don't speed up with the override, so past a point the path is the lever left.

On a factory floor that adds up. At 66.88 s the robot does about 53.8 cycles an hour. At 64.76 s it does 55.6. That is roughly 14 more cycles, or about 56 more bottles, per 8-hour shift from the same robot at the same speed. Running faster adds wear and cuts safety margin. Saving time through the path avoids both.

Limits

Energy was not measured. Shorter cycles usually mean less energy per part, but here that is a likely benefit, not a result. I'm also not sure the 50% run was in T2 (T1 caps speed at 250 mm/s), so treat that number as indicative. Each run was timed once, by hand.

Feedback and next steps

The lab assistant pointed out that an in-program timer ($TIMER) would be more precise than a stopwatch. I'd use it next time. I also left per-move velocities at baseline values. Raising them to 100% on the optimised path should cut the time further.

What I took away

Speed isn't the only lever. Once speed stops paying off, path design wins. Change one thing at a time, and know your measurement error before you claim a result. Fixtures matter a lot for a taught robot.

This was also my first hands-on time with a KUKA KR robot, after a CAD and PLM background in CATIA and 3DEXPERIENCE. Taking a design from CAD into a real robot cell was the most useful part. The full write-up is on the project page, and the CAD and results are on GitHub. You can find me on LinkedIn.

If you work with KUKA or cycle-time optimisation, I'd like to hear how you'd approach the velocity and timer changes.


Quick answers

Can you cut robot cycle time without increasing speed?
Yes. On a KUKA KR10 pick-and-place task, moving approach and retract points closer, placing via points by hand and removing a wasted axis rotation cut the cycle from 66.88 s to 64.76 s at the same 100% speed.
How much did path optimisation save?
2.12 s per cycle, or 3.2%. That is about two-thirds of the 3.1 s gained by going from 75% to 100% program override.
Was energy measured?
No. The NI DAQ setup was not ready, so cycle time was compared instead. Shorter cycles usually mean less energy per part, but that is a likely benefit here, not a measured result.
← Back to all blogs Share on LinkedIn ↗