Collaborative Robots: Mobile and Adaptable Labmates (hero)

Collaborative Robots: Mobile and Adaptable Labmates

The scale and versatility of ‘cobots’ signal a revolution in high-throughput applications, from drug discovery to COVID-19 testing

Brandoch Cook, PhD

Brandoch Cook, PhD, is a freelance scientific writer. He can be reached at:

ViewFull Profile.
Learn about ourEditorial Policies.

Collaborative robots, “cobots” for short, promise a marked improvement compared to legacy industrial robotics in worker safety, human complementarity, and sleek aesthetics. Colgate, Peshkin, and colleagues originally designed cobots as Intelligent Assist Devices under the aegis of General Motors to improve assembly line safety. In the ensuing years, cobot technology has evolved to incorporate lightweight and contamination-resistant alloys and synthetics, smaller footprints and increased mobility, and improved haptic interfaces to allow intuitive communication between themselves and their human counterparts. Consequently, cobots have begun a steady migration from the factory floor to the laboratory space.

Although they do not seem to share many immediate or short-term goals, manufacturing and laboratory science are subject to some of the same mundane key performance indicators, which can be summarized most simply as a directive to maximize throughput and minimize cost while maintaining consistency. In factory environments, these metrics have been optimized for half a century by supplementing the human workforce with industrial robotic systems to handle repetitive, high-payload tasks where machines can mitigate stress and produce more uniform parts to higher regulatory standards. These principles have carried over into lower-payload tasks in scientific research, particularly in chemical screening workflows for early-stage drug discovery, where automation of liquid handling, chemical deposition, and library curation has long been de rigueur. However, industrial robots can make unsuitable research assistants. The most obvious discrepancy between the auto plant and the lab bench is the latter’s constraint on available space. Additionally, industrial robots are typically one-task specialists that function separately from each other and require extensive individual preprogramming before changing operations. Therefore, industrial arrays in the laboratory have largely been consigned to dedicated core facilities run by designated experts. The necessity for guarding mechanisms in industrial robots to prevent unintentional human contact and potential injury greatly expands their footprints and precludes their extensive use in finite research spaces.

In comparison to industrial robotics, cobots feature rounded edges with smoother and more pliable surfaces to mitigate pinch-point contact injuries. Moreover, innovative design features include impact sensing and force- and torque-limiting mechanisms to identify human contact at the point of initiation and respond by immediately stopping, rather than following through with unimpeded force. Contemporary cobots are thus beginning to function at the apex of true responsive collaboration, with real-time interaction and teaching mode adjustment by users whose expertise is in scientific protocol rather than robotic design and machine learning, and who work directly in contact with active robotics. 

Because of these developments, screening laboratories can begin to leverage automation to deliver its most promising aspects to the sciences: 1) to take the tedious repetition of low-payload tasks that comprise initial discovery processes, remove them from human hands, and place them in a system that is inherently  more consistent, quantitative, and reproducible; 2) to circumvent the circadian rhythm, screening 24 hours per day without pause to streamline and expand the discovery process so that investigative teams can vet many more drug candidates, identify more putative hits, and eliminate more false positives through iteration; and 3) to utilize decreased footprints and diversified end-of-arm tooling to create greater adaptability. The resulting ability to streamline and standardize smaller-scale tasks such as tissue culture incubation and maintenance may allow for discovery of more effective drugs using more powerful automated genetic and phenotypic screens with complex three-dimensional tissue culture models.

Thermo Fisher Scientific offers the inSPIRE collaborative automation platform. Its modular design can accommodate a variety of automated units, centered around the Thermo Spinnaker microplate cobot, which uses vision-based teaching to learn plate placements, and functions in a 360-degree perimeter where it handles barcoding, container orientation, and lidding or de-lidding operations. Accessory collaborative arrays are wired via SmartShelves and activated or shut down via SmartHandles, which indicate system status and health by color and provide haptic run status feedback to the user. Multiple inSPIRE platforms can be linked to optimize complex parallel processing workflows, and can additionally be coupled to Thermo’s Vanquish series of liquid chromatography instruments to create an automated analytics workflow and obtain read-outs for proteomic effectors of novel therapeutic targets.

The CoLAB system from HighRes Biosolutions promises to maximize workflows in screening and discovery to augment and improve pharmaceutical pipelines, and provide a cost-control measure in the pharma industry, which is currently beset by escalating costs of up to $3 billion to bring a novel drug from discovery to market. Through a partnership with AstraZeneca (AZ), HighRes has developed CoLAB automation around a re-branded KUKA seven-axis arm initially designed to collaborate with astronauts whose spacesuits limit their abilities to manipulate objects in zero gravity.

The AZ CoLAB array can screen approximately 300,000 compounds in a 24-hour day. For comparison, your humble author has also successfully conducted a high-throughput chemical screen, in which he assayed approximately 48,000 chemicals in just under five months. With the obvious improvement in throughput and consistency, the HighRes/AZ collaborative expects to screen up to 50 million compounds annually. Moreover, this early-stage discovery research is arranged under an open innovation program through the UK Centre for Lead Discovery, with facility access provided to strategic partners such as Cancer Research UK and Charles River Laboratories. With this approach, HighRes and AZ can leverage the maximum potential for the differential research roles filled by cobots and their human counterparts: on the one hand, to perform tasks at a scale and reproducibility of which humans are not capable, and on the other, for scientists to design and program actions required to complete studies, and create subsequent strategies based on data obtained and integrated through automation.

According to Jon Mole, Northeast sales manager, the mission of HighRes is “vendor-, device-, and application-agnostic integration” of robotics systems in a manner that works with clients to “understand what their science and business goals are and tailor a solution specific to those goals…by modeling workflow and predicting throughput.” Therefore, HighRes commonly builds arrays around the KUKA arm using their MicroStar platforms, or around the PreciseFlex PF400 arm (rebranded as the ACell) as a “robot on a rail,” and integrates them into systems customized for pharma and biotech clients, academic screening centers, and individual labs. Beyond these foundational units, the sky is essentially the limit for what devices can be integrated, with more than 250 currently incorporated in various extant HighRes systems. With the inherent modularity, it is possible to begin small, and then vertically integrate by stacking devices and maximizing capability in restricted footprints common to academic laboratories. In addition to the usual force-limiting mechanisms inherent in other cobots, HighRes also employs a series of laser sensors to create a detection field of approximately two meters beyond the robot. If a human is outside the zone, it functions at full industrial speed, and if a human steps inside the zone, it slows down to collaborative speed. By extension, Mole suggests that the only devices that should still be integrated via industrial robotics are those handling operations requiring sterility or chemical safety, because the human must, by necessity, be outside the enclosure anyway.

HighRes is unique within the collaborative robotics industry in its long-standing relationship with clients, in which systems have been augmented and re-purposed over more than 10 years of use and optimization. Screening centers and core facilities at the Rockefeller University  and the Broad Institute are examples of the success of this approach. Investigators at the Institute for Systems Genetics at NYU Langone have used HighRes collaborative builds to enable their genetic engineering projects. Strikingly, the group has more recently expanded their arrays via CoLAB Flex, a system in which devices are mounted on mobile carts, and pivoted their use toward high-throughput, low-cost COVID testing for New York City residents.

Bill Gates once remarked that “the first rule of any technology used in a business is that automation applied to an efficient operation will magnify the efficiency. The second is that automation applied to an inefficient operation will magnify the inefficiency.” If science follows his axiom, it stands to reason that there are operations that are intrinsically, or by design, inefficient and therefore should not be automated. Although cobots promise to streamline many predictable-outcome, smaller-payload, low-volume tasks that are outside of industrial robotics’ purview, the corollary to this proposition is that there are scientific tasks that are too small, qualitative, and unpredictably analytical to automate even collaboratively, and that attempts to do so would result in chaotic and superfluous downstream readouts. Therefore, laboratory managers and leaders interested in investing in cobot systems must perform cost-benefit analyses unique to their individual research missions, and should begin with simple layouts, gauging their utility over time before deciding to scale, either upward or outward.