Cooler-box depalletizing (NALO Plainfield)

How we'd approach it

The customer has already handed us the cell layout, so the decomposition follows their stated workflow rather than anything we'd invent: an AMR drops a full pallet of ~25 lb Orion preconditioned coolers into a fenced cell, and a mobile manipulator lifts each box off the stack onto a conveyor at ~10 mm placement. Read as primitives, the placement half is the easy half — 10 mm onto a moving belt with high failure tolerance is forgiving, and it's squarely the kind of put-down our soft-body work already covers. The hard half is acquisition. The customer was explicit that open box flaps defeat vacuum grippers, which means the default depalletizing tool is off the table and the whole task pivots on finding an alternative grasp — pinch from the sides, scoop from underneath, or a hybrid end-effector — that can secure a 25 lb load whose top face is uncooperative. Layered on top is the side-reach problem flagged in our internal note: a full pallet means the manipulator works from a floor-level bottom layer up to a shoulder-height top layer, and the reach envelope changes with every layer removed. So the approach is really three questions stacked — how to grip a flap-open box, how to reach across a depleting pallet, and how to do both inside whatever cycle time 200/hr implies. We don't actually know the cycle budget the robot must hit: 200/hr is the human rate, but is the robot expected to match it on day one, or is a slower depalletizing rate acceptable if it removes the ergonomic exposure? That answer reshapes the hardware choice.

Why it's landable for us now

The roadmap reading is more encouraging on the manipulation than on the mechanism. The one primitive this task is tagged against, soft_body_handling, sits at works_in_demos with a roughly three-month horizon and rough confidence — Demo 2's mixed-fragility compliance is core to exactly this, so the act of compliantly grasping and placing a semi-rigid, flap-bearing box is demonstrable, if not yet pilot-proven. That alone would make this close. The honest caveat is that the gating difficulty here isn't captured by any roadmap primitive: the vacuum-gripper blocker and the side-reach challenge are end-effector and platform questions, not manipulation-intelligence gaps, and the roadmap doesn't tell us where we are on either. Our own internal note already names this — soft_body is solid, but the open-flap and side-reach problems push the task to a stretch band — and that read is consistent with the roadmap rather than contradicted by it. So the verdict is that the cognitive part of the task is roughly a quarter away while the cognition is rough-confidence, but the deployment is gated on a grasp-and-reach solution whose maturity we can't yet trace to a capability state. That's also why I'd want to know the real loaded-box weight and the manipulator payload we'd bring: the customer says ~25 lb, but is that confirmed for the heaviest preconditioned cooler, and does a mobile platform that can both reach the full pallet envelope and lift that load with a non-vacuum gripper exist in our hardware plan, or is it something we'd have to spec?

Fit against the baseline

The interesting thing about this seed is that it doesn't behave like its vertical. Lilly is pharma_lab, and the baseline puts Pharma/Lab Automation in Tier 2 — ±0.5-1mm tolerance, medium forgiveness, a 12-18 month sales cycle, Phase 3 capabilities required. This task fits none of that. A 10 mm placement with high failure tolerance, handling a deformable-flapped box off a pallet onto a conveyor, is functionally a logistics depalletizing task wearing a pharma badge: it lines up with the baseline's most forgiving rows — E-Commerce/Logistics at ±5mm, very high forgiveness, or Agriculture-style ±5-10mm — and with Phase 1-2 capability, not Phase 3. So against the vertical's general position this task is markedly more favorable: it's a Tier-1-shaped problem inside a Tier-2 account. That's a strategically attractive shape — it's the 'adjacent industry pilot first' logic from the baseline, an account that will tolerate a robot at modest success rates because the alternative is a documented top ergonomic strain (Lilly's own Advanced Tool score of 44, with shoulders, neck and elbows flagged). The medium tactile value the customer assigned fits too: this isn't a bruise-prevention produce problem, the tactile job here is grip security on a heavy box, not delicate force modulation. The open question the baseline raises is whether landing a forgiving-tolerance depalletizing task actually earns us anything toward the harder pharma work we'd eventually want in this account — autoinjector and kit assembly — or whether it's a standalone logistics win that happens to live under a pharma logo. Worth knowing whether Lilly sees this as the front door to a broader robotics relationship or a one-off ergonomic fix.

Open questions

• Is the robot expected to match the human's ~200/hr, or is a slower rate acceptable since the goal is removing ergonomic exposure rather than adding throughput?
• What does 'open box flaps' mean physically — are flaps always open and floppy, can they be grasped or pushed aside, and is there a side or under-box surface a non-vacuum gripper could purchase on?
• Is ~25 lb the confirmed weight of the heaviest loaded cooler, and what manipulator payload would we need to carry that with a non-vacuum end-effector?
• Does the cell require a genuinely mobile manipulator to cover the floor-to-shoulder pallet reach, or can a fixed-base arm at the conveyor handle the envelope?
• Where is our actual maturity on a flap-tolerant grasp mechanism — this is the true gate and it isn't reflected in any roadmap primitive?
• Does landing this forgiving-tolerance depalletizing task open a path to the harder Phase 3 pharma work in this account, or is it a standalone logistics win under a pharma logo?
• Is the placement genuinely onto a moving conveyor, and does belt motion add any timing constraint beyond the ~10 mm spatial tolerance?