Vision picking startup allocation without a product buffer

Two robots can share a picking conveyor successfully in normal production yet lose products during startup. The downstream robot may receive targets before an indexed tray provides somewhere to place them. Without a physical product buffer, changing target allocation helps only when the upstream robot has both enough capacity and enough permitted empty pockets. This guide shows how to plan that startup phase, limit product admission, and change allocation without losing track of trays or targets.

Decide what upstream-only picking can actually solve

Use upstream-only allocation during startup only when the first robot can complete every admitted pick into an authorized destination. Do not interpret an idle downstream robot as permission to fill its assigned pockets upstream.

There are two different startup strategies:

• Preserve the normal pocket split. The upstream robot fills its usual pockets. Reduce incoming product flow until the downstream station has eligible trays.
• Temporarily change pocket responsibility. The upstream robot fills additional pockets under a validated startup recipe. Track those occupied pockets so the downstream robot cannot fill them again.

The first strategy is generally simpler. The second requires verified reach, compatible placement geometry, additional cycle capacity, and explicit tray occupancy records.

Neither strategy creates storage. If incoming products exceed available picking and placement capacity, delay or reduce product feed. If feed cannot be controlled, define an approved reject or production-stop response before commissioning.

For problems caused by an indexed station stopping on its final position, use the separate guide to diagnosing final-item stops in indexed work areas. That is a different fault from allocating products before destinations exist.

Track trays and pockets separately from detected products

A camera target identifies something available to pick. It does not prove that a tray is present, correctly indexed, or ready to receive that product.

Maintain a tray record that follows each physical tray through the indexing sequence. At minimum, record:

• Tray identity or a validated logical identifier.
• Current station and confirmed index sequence.
• Recipe and expected pocket pattern.
• Occupied, available, reserved, and uncertain pockets.
• Which station is responsible for each pocket group.
• Completion status and any quality hold.

Keep product targets in a separate record containing detection identity, conveyor position, ownership, and expiration status. Connect a product target to a placement reservation when the application commits to the operation.

For startup planning, define placement readiness as the conjunction of several conditions: the correct tray is present, indexing has finished, the pocket is available, and the station can execute the validated placement cycle. Use the actual machine's documented interfaces to implement those conditions.

A placement command is not confirmation that a pocket became occupied. Update occupancy from the application's validated completion criteria. If placement outcome is uncertain, mark the pocket uncertain and stop automatic reuse until the uncertainty is resolved.

Use explicit startup states instead of a timed override

A timer can estimate when a tray should arrive, but it should not decide that the tray has arrived. Use confirmed tray movement and destination readiness to advance the allocation state.

These are application-level states, not replacement safety functions. Protective devices, interlocks, and safety-rated control remain independent of the allocation logic.

Do not release normal production merely because the downstream robot reports available. It also needs a valid tray, remaining pocket capacity, and enough tracking time to complete its assigned picks.

The related guide to conveyor vision startup synchronization addresses acquisition and tracking alignment. Establish that alignment before evaluating the allocation policy.

Calculate the startup admission limit

Start with complete pick-and-place cycles, not robot motion speed alone. Include approach, gripping, travel, placement, confirmation, and the time unavailable during indexing or coordination.

For an initial estimate, compare the cycle limit with the rate at which authorized pockets become available:

These are planning bounds, not scheduling guarantees. Bursts, product spacing, indexing pauses, and conveyor deadlines can require a lower operating rate. Use measured cycle distributions rather than relying only on an average.

Worked example: two index intervals before downstream readiness

Assume an illustrative application with these conditions:

• Each tray has 12 pockets assigned to the upstream station and 12 assigned downstream.
• The tray conveyor indexes every 8 seconds during the planned sequence.
• The first eligible tray reaches the downstream station after two index intervals.
• Each robot completes a representative pick-and-place cycle in 0.50 seconds.
• The normal incoming product rate is 3 products per second.

The estimated startup waiting period is:

At the normal feed rate, 16 seconds introduces 48 products. The planned upstream trays provide only 24 authorized pockets during that interval. Even though the upstream robot has a theoretical cycle capacity of 32 picks, it still lacks destinations for 24 incoming products.

Reducing admission to 1.5 products per second introduces 24 products over the same interval. Each upstream group requires approximately 12 × 0.50 = 6 seconds of service within an 8-second index interval. That leaves a nominal 2-second allowance, which must cover the actual coordination requirements and measured variability.

After downstream readiness is confirmed, each station has a pocket-rate bound of 1.5 products per second. Their combined bound is therefore 3 products per second, subject to successful scheduling and suitable margin.

This example assumes normal pocket responsibilities remain unchanged. It also assumes the planned index sequence is achievable. If incomplete placement prevents indexing, recalculate from the actual sequence rather than forcing the conveyor to maintain the estimate.

Check conveyor deadlines as well as average rates

A finite picking belt can provide transit time, but it is not an unlimited buffer. Every target has a last usable picking position.

Estimate the remaining opportunity for each target:

For example, 600 mm of remaining usable travel at 300 mm/s gives approximately 2 seconds. A downstream placement delay lasting 16 seconds cannot be solved by retaining that target in a software queue. The physical product will leave the usable region first.

Evaluate the deadline at the relevant robot's picking window. Camera-to-robot travel can delay the arrival of newly admitted products, so feed changes must be scheduled with that transport delay in mind.

The guide to camera placement relative to conveyor-tracking robots explains that geometry. For flow regulation after startup, use the separate discussion of vision queue conveyor speed control.

Change allocation without changing existing ownership

Use a controlled boundary between startup and normal allocation. A mode bit that changes asynchronously while several consumers request targets can create ambiguous ownership.

A practical transition sequence is:

1. Confirm the downstream tray identity, index completion, and available pocket group.
2. Calculate the downstream capacity available during the first normal allocation interval.
3. Temporarily restrict new admissions or commitments if the transition requires it.
4. Record an allocation version and a target sequence boundary.
5. Allow already committed targets to finish under their existing ownership and destination reservations.
6. Apply the new policy to targets beyond the agreed boundary.
7. Confirm that both stations and the allocator report the same allocation version before raising feed.

Do not copy an existing target into another robot's queue. If the documented system supports reassignment, require an explicit release and acknowledgment before another consumer can claim it. A target already being executed should not be reassigned through ordinary allocation logic.

When upstream filling temporarily includes downstream pockets, the transition also needs pocket-level accounting. The first arriving trays may contain fewer available downstream positions than later trays. Ramp admission according to actual remaining capacity, not the normal tray recipe alone.

For spatial ownership questions, use target routing across overlapping work areas. The startup policy described here decides when allocation becomes eligible; spatial routing decides where an eligible target can be handled.

Select integration hardware around observable interfaces

The allocation policy belongs in the documented application and control interfaces. A camera detects products, while the application combines detections with tray state and production readiness.

When reviewing vision controllers, check supported camera acquisition, communication interfaces, software compatibility, event logging, and configuration persistence. Do not assume that selecting a controller also supplies a particular robot allocation function.

The MV-VC3301H-128G66 is a controller candidate to review against those integration requirements. Confirm the application software and supported interfaces for the intended installation.

The MV-VC2140-128G40-1T is another controller candidate. Compare the actual connection and software requirements before choosing between platforms.

For moving products, review area scan cameras against exposure, shutter behavior, trigger timing, and acquisition throughput. The MV-CU020-90GC uses a global shutter and GigE interface. Those characteristics do not establish application detection accuracy or timing by themselves.

Use the camera finder to narrow the acquisition hardware. Keep camera selection separate from the proof that the destination and allocation logic are correct.

Log detection, claim, placement reservation, placement completion, index confirmation, and allocation changes with a shared timing reference. The resulting event sequence should explain why every admitted product received its particular owner.

Commission with a short, measurable checklist

Test the state transitions at reduced production demand before attempting the normal feed rate. Use the machine's approved commissioning procedures and keep protective functions active.

• Verify the initial tray map. Confirm that the logical tray identifiers match physical trays throughout the first index sequence.
• Start without downstream destinations. Confirm that the downstream robot receives no unsupported commitments while its tray is absent.
• Measure upstream-only throughput. Count confirmed placements over complete index intervals, including pauses and coordination delays.
• Test a delayed tray. Confirm that a late index confirmation prevents the allocation transition rather than merely delaying a timer.
• Inspect the first downstream tray. Compare physical occupancy with every recorded available and reserved pocket.
• Record the transition boundary. Verify that targets on either side receive the intended policy without duplicate ownership.
• Check the first normal interval. Confirm that available downstream capacity supports the increased admission rate.
• Test a failed placement outcome. Confirm that an uncertain pocket cannot silently return to the available pool.
• Account for all products. Reconcile admitted, placed, rejected, and still-tracked products at the end of the test.

A useful reconciliation is:

Compare that physical balance with the target ledger. Duplicate detections or unconfirmed removals must not become apparent spare capacity. For image-overlap problems, use duplicate detection checks for overlapping conveyor images.

Recognize when upstream-only allocation is the wrong answer

Upstream-only allocation cannot solve a destination shortage if the upstream robot fills its permitted pockets faster than new trays arrive. Increasing robot speed does not change that pocket supply.

Temporarily filling every reachable pocket can also move the shortage downstream. A tray arriving already full provides no placement capacity, even when the downstream robot is ready. This can produce alternating periods of starvation and overload unless admission follows actual occupancy.

Holding a product in a gripper is not automatically an acceptable buffer strategy. Product retention, permitted hold time, process quality, and recovery behavior all require validation. Do not make startup depend on an unverified extended hold.

If feed cannot be reduced, destinations cannot be added, and products cannot remain within the picking windows, allocation logic alone is insufficient. The process needs a defined alternative: controlled feed interruption, an approved reject route, or a revised production sequence.

A later interruption is also not simply another cold startup. Existing reservations and products may remain in flight. Use the separate guide to vision picking restart handshakes when recovery must preserve those commitments.

Turn the startup plan into a reviewable specification

Before ordering hardware or changing the application, document the tray sequence, pocket responsibilities, startup admission limit, transition boundary, and response to missing readiness. Include measured cycle times and conveyor deadlines rather than only a desired production rate.

Use the camera configuration starting point for acquisition selection. For integration review, contact applications support with the tray map, timing diagram, and required interfaces. The key decision is whether admitted products have both an available robot and a valid destination throughout startup.