Retrofits Beyond Factory Specification¶
Step 4 · Extend: How to go beyond the catalog?
Summarizes Section 5 (Customizing Platforms Beyond Their Factory Specification) and Table 5 of the review.
The optional gripper arms on the Opentrons Flex and the Hamilton Microlab STAR are the manufacturer-approved route to generalization. A larger body of work gets there without the vendor. Researchers keep a platform's gantry, deck and control loop, then swap its end-effector, add a sensor or replace its control software to unlock a capability the manufacturer never sold.
The literature scenarios ask how a material should be handled. This work asks a different question: which layer of a platform you already own should you modify (mechanical, sensing or software) so that one machine can cover more steps of a formulation workflow? Table 5 groups ten studies by that layer. Three of them also appear among the liquid-handling scenarios; seven are new here.
Table 5: Retrofits that extend platforms beyond their factory specification¶
| Layer modified | Base platform | Novel idea | New capability unlocked | Ref |
|---|---|---|---|---|
| Control parameters (unmodified hardware) | Opentrons OT-2 | Bayesian-optimize aspirate/dispense speed against gravimetric feedback instead of a fixed viscosity lookup table | <5% error dispensing liquids up to 1275 cP on a robot rated for water-like liquids | [11] |
| Opentrons OT-2 | Treat the pipetting robot’s own dispense-time/mass data as a viscosity-proxy signal | Turns a liquid handler into a high-throughput viscometer with no dedicated rheometer | [23], [81] | |
| End-effector replacement | Generic robot arm | Replace the pipette tip with a vibrating acoustofluidic coupler | Mixes liquids ~10⁶× more viscous than water without a pump | [25] |
| Opentrons OT-1 | 3D-printed adapter turns the robot’s own stepper motor into an infuse/withdraw syringe pump | Extends a microliter-rated robot to 15–50 nL transfers for under $6,000 total system cost | [82] | |
| Councill’s modified Opentrons OT-1 | Add two-position/selector valves to the same syringe-pump retrofit | Same platform doubles as an autosampler feeding an LC-MS instrument | [83] | |
| Generic robot arms | Sub-$200, three-part 3D-printed pipette end-effector | Lets arms that cannot grip a pipette at all reach commercial-liquid-handler accuracy | [84] | |
| Generic tool-exchange gripper | Container-shape taxonomy drives dual-material compliant gripper-finger design | One gripper reliably grasps vessel shapes a stock gripper’s fingers cannot | [85] | |
| Sensing add-on via tool-changing | Opentrons OT-2 | Adapter with two mounting bores lets the existing tool-change slot carry a pipette or a camera | Same robot alternates dispensing and in-situ imaging without a separate imaging station | [86] |
| Opentrons OT-2 | Mount a camera + transilluminator, pair with a Detectron2 vision model | Liquid handler gains automated colony picking (~240/h) at a fraction of a dedicated picker’s cost | [87] | |
| Software/control layer | Opentrons, Hamilton, Tecan | Hardware-agnostic Python driver replaces vendor firmware | Exposes physically possible moves vendor software never programmed, e.g., dual-plate pipetting | [88] |
Three families of retrofit¶
1. Replace the end-effector¶
The most direct retrofit keeps the motion system and replaces the end-effector (Table 5, rows 3–7). There are two threads.
Reuse the robot's own actuator. Councill et al. turn the stepper motor of the now-discontinued Opentrons OT-1 into a nanoliter syringe pump [82]. Axtell et al. show that the same retrofit, with valves added, also works as an LC-MS autosampler [83].
Platform-agnostic tools. Yoshikawa et al.'s open-hardware digital pipette [84] and Zwirnmann et al.'s gripper fingers, designed from a container-shape taxonomy [85], fit any compatible arm. They give up deep integration in exchange for reuse across a fleet.
2. Use the tool changer as a sensor mount¶
A second family uses the tool-change mechanism, built for pipette tips, to mount a different tool. Nazeri et al. [86] and del Olmo Lianes et al. [87] both mount a camera on an Opentrons OT-2: the first to monitor dispensing, the second to pick colonies. What is new is not the commodity camera but the idea of treating a consumables interface as a general-purpose mount.
3. Replace the control software¶
A third family works at the software layer. Wierenga et al.'s PyLabRobot [88] replaces proprietary control software with one hardware-agnostic Python interface for Opentrons, Hamilton and Tecan platforms. It exposes moves that are physically possible but that vendor firmware never programmed, such as an 8-channel head servicing two microplates in one motion.
Layers like this underpin every other customization. A robot cannot act on a new sensor through firmware that refuses non-standard commands, and a formulation workflow cannot be orchestrated across instruments that refuse a common interface. See also Orchestration.
Why it matters¶
Together, these retrofits can turn a $15,000–30,000 liquid handler into a nanoliter dispenser, an autosampler, a colony picker or an imaging station. The cost is a few hundred dollars in printed parts plus a workaround in the control layer. They are a lower-cost counterpart to the new-hardware scenarios E–H in Figure 3, and they turn single-purpose robots into the multi-function platforms formulation SDLs need.
Every new capability is a new failure mode
Proving a retrofit on the physical cell is slow and risky, which is the motivation for rehearsing it first in simulation and digital twins.
See also: Physical Reinvention · Integrated Platforms