Robotics · End-of-line packaging
The Truck Test
A palletising cell that plans every place for a cobot on a lift column, then brakes the stretch-wrapped pallet it built in a physics simulation.
Overview
A palletising cell built in simulation, for 48 cases (400 × 200 × 300 mm, 8 kg each) on a Euro pallet. The cases go on in one of two patterns: column, where every layer is the same, or interlocked, where every second layer is turned 90 degrees. The robot is a six-axis cobot on a lift column. Our own inverse kinematics solves every place pose on both pallets in closed form. The column raises the arm one layer at a time, and each case goes in on a straight-line approach into its corner. The film shows one lift move and one place.
The finished pallet is stretch-wrapped and braked in a MuJoCo physics simulation. We raise the deceleration in 0.1 g steps on the same load, following the step logic of the EUMOS 40509 test, until a step fails. A step fails if any case moves more than 60 mm (5 % of the load height) while the load is braking, or is still more than 48 mm (4 %) from where it was placed once the load settles. These limits are the stricter of two published summaries of the standard's criteria; we did not read the standard itself. The film is modelled as pre-tensioned bands plus tension-only ties fixed to the outer cases. We set the band tension from a stretch-wrapper maker's recommended minimum containment force for a heavier load, so the tension is an upper estimate.
In this model the column stack passes 0.5 g and fails at 0.6 g. The whole load slides forward on the deck and again between the layers, and a front case in the top layer ends up 57 mm from where it was placed. In the interlocked stack, the cases in each layer stay together, and it passes every step up to 0.8 g. At 0.6 g its largest shift is 45 mm. The margins are narrow. With the other summary's 60 mm limit, the column stack would also pass 0.6 and 0.7 g, and the interlocked stack passes 0.8 g by under 1 mm (47.1 mm). Read the pass levels as results of this model with these limits. Interlocking is not free. In pallet-load compression tests at RIT for the Fibre Box Association, with their boxes and patterns, interlocked stacks lost 49–54 % of their compression strength, against 12–20 % for column stacks. We cite that result; we did not simulate compression. The cases are rigid boxes and the film is a simplified model. The friction and film stiffness are assumed values, not measurements.
What we built
- A layer-pattern generator for one case size on a Euro pallet, in column and interlocked patterns.
- Closed-form inverse kinematics for a six-axis cobot, plus a seventh, vertical axis: the lift column. Every place and approach pose on both pallets is solved, checked by forward kinematics and kept clear of the elbow and wrist singularities. Moves are straight lines into and out of each pick and place, with joint-space transits between them.
- A MuJoCo model of the wrapped load: 48 rigid cases with friction contacts, and the film as 12 pre-tensioned bands plus 344 tension-only ties.
- A stepped braking test on the same load: a 0.1 s ramp, 0.8 s at the step's deceleration, a 0.1 s ramp down and 0.6 s to settle. The limits are stated above.
- Checks of the contact model against hand calculations: one case slides on the deck at its friction coefficient, and a free-standing column of cases tips at its width-to-height ratio.
Tech stack
Python · MuJoCo · NumPy · three.js
Numbers
| Number | How we measured it |
|---|---|
| 0.6 g | The first step the column stack fails, in our MuJoCo braking simulation. The steps rise by 0.1 g on the same wrapped load, and the highest step it passed is 0.5 g. At 0.6 g a top-layer case is still 57 mm (limit 48 mm) from where it was placed once the load settles. Under the other published summary's 60 mm limit it would pass 0.6 and 0.7 g. |
| 0.8 g | The highest step the interlocked stack passes in the same simulation, with the same cases, film and limits, by 0.9 mm (47.1 mm against 48 mm). It fails at 0.9 g, where a case ends up 58 mm out of place. At 0.6 g its largest shift is 45 mm. |
| 49–54 % | Loss of compression strength for interlocked stacking, against 12–20 % for column stacking. Cited from pallet-load compression tests at RIT for the Fibre Box Association (2021), for their boxes and patterns; not simulated. Our interlock is a 90° cross-stack. |
