A palletising cell has a rated throughput, and most cells do not reach it. The gap is rarely the robot’s speed – it is the path length, the number of approach and retreat moves, and the time spent correcting a box that did not land where it was aimed. All three are functions of the pallet pattern and of how repeatable the boxes are, and both are decided before the robot arrives.
Travel distance dominates
Each cycle is a pick, a move, and a place. The move is the largest single component and its length is set by where the box is picked from and where it is placed. A pattern that alternates between two distant corners of the pallet doubles the average move compared to one that fills in a sweep. Where throughput is short, the first thing to look at is the place sequence rather than the robot speed setting, and it is a programming change rather than a capital one.
Approach and retreat height
The robot cannot travel at the pallet surface; it has to lift clear of what is already placed and then come down. That adds two vertical moves per cycle, and their length depends on the height of the stack already built. A tall pallet costs more per placement at the end than at the beginning, which is why throughput is not linear in boxes per pallet. Where the pattern can be arranged to keep the working height lower for longer – wider layers rather than taller stacks – the average cycle improves.
Re-grips and corrections
Every time a box does not land true, the cell either accepts a poorer pallet or spends time correcting. A pattern that requires tight placement needs repeatable boxes and a well-tuned gripper; a pattern with generous gaps tolerates variation and runs faster. The trade is between pallet quality and rate, and it is worth making explicitly rather than discovering it as a cycle-time shortfall. The tolerance side of this is in gripper faces and box side walls.
Layer pads and the extra pick
A pattern with layer pads between layers adds a pick-and-place cycle per layer, and those cycles carry no product. Where pads are needed for stability they are worth it; where they have been included by habit they are pure cost. The alternative is usually a box with a base that interlocks or a wall stiff enough to carry the stack without a pad, and that is a container specification question rather than a robot one.
Mixed sizes in one pallet
Mixed-size palletising is slower per box than uniform palletising, because the pattern has to be computed per layer and the placements are less repeatable. Where a line produces mixed pallets, the realistic rate is lower than the cell’s rated rate, and it should be quoted that way. Where the mix can be reduced – fewer sizes, or sizes that share a module – the rate improves for free. This is pursued further in mixed-case palletising and box consistency.
Where the box helps
A box with a consistent footprint, a stiff base and square walls places accurately and needs no correction. A box with a bowed wall, a variable footprint or a base that deflects places approximately, and every placement costs a little extra time. Over a shift that difference is large, and it is bought with container specification rather than with robot capability. It is usually the cheaper of the two.
How to measure it before committing
Time twenty consecutive placements on the real pattern with the real boxes, and separate the move time from the approach time from any correction time. The breakdown shows where the rate is going, and it nearly always shows that one of the three is larger than anyone assumed. Send us the pattern and the target rate and we will tell you which box construction supports it.