The future of intelligent lab orchestration depends on one critical foundation: hardware that can operate reliably without human intervention.
Here at HighRes, we have been working on our new Perception suite: a set of robot- and system camera-vision tools built into our automation systems (Video 1). The picture it paints of where lab automation is heading is hard to ignore.
Video 1. The HighRes FlexPod with robotic Perception system brings a new level of intelligence to laboratory automation through three key capabilities: (1) Rapid deployment through auto-teaching, (2) Intelligent quality control and verification, and (3) Dynamic protocol adjustment.
The framing matters here. The team wasn't pitching Perception as a novelty feature. They were explicit: our lab automation systems are already reliable. The goal is to push that reliability even further, to turn a system lab scientists occasionally babysit into one you can genuinely walk away from.
Intelligent labs require the system to see, interpret, and respond to the unexpected, like the mislabeled plate, the obstacle left in the wrong place, or the lid that got left on when it shouldn't have been.
The range of applications Perception can cover is broad, but here are some of the highest-value use cases where it tends to make the biggest difference.
Before a run begins, Perception can confirm that everything is where it should be. That means checking whether a liquid handler deck is configured correctly, how many pieces of labware are loaded in a stacker, or whether consumables are oriented properly. Simple checks that, when missed, can derail an entire run.
A lot of automation errors are foreseeable (e.g., a plate left in a device from a previous run, a finicky door that doesn't open all the way). Perception can check for these conditions proactively. A labware presence/absence model can confirm a nest is clear before a plate is placed there, avoiding crashes. Physical conditions that might cause downstream issues can be flagged before the protocol ever reaches them.
Not all labware checks are the same. Perception handles three distinct kinds: confirming the right labware has been loaded for the run (identify), verifying that a plate's lid or seal has been properly removed before it moves to the next step (qualify), and counting tubes or tips in a rack and feeding that positional data back to the system so it knows which consumables to use or skip (quantify).
This is where Perception shifts from detection to action. Checks can be inserted anywhere in a workflow, whether tied to protocol operations, run order, or specific resources. When something unexpected turns up, the system doesn't just stop and wait. That plate left in the wrong place? It gets automatically moved to a quarantine location and the protocol continues. A plate that needs to go into an instrument de-lidded? The system handles it. Of course, simpler responses are always available, too. Throwing a system error or warning when a check fails is still an option when that's the appropriate response.
The models that power Perception — classification, anomaly detection, object detection, quantification — can be combined or trained from scratch to fit a lab's specific needs. The out-of-the-box capabilities are a starting point, not a ceiling.
One of the more practical strengths of the Perception solution is that it isn't tied to any specific camera, sensor, or robot platform. It's designed to work with a range of system and robot cameras, including small form-factor cameras embedded directly in stackers, cameras built into robotic arms, and larger pan-tilt-zoom cameras mounted around your system. In the demo, all three configurations appeared in the same workflow, each serving a different detection purpose.
This flexibility extends to existing infrastructure as well. HighRes can work with you to evaluate whether your current camera systems and video setup are compatible with Perception, though not all hardware will qualify, as certain technical requirements must be met. For labs that are already equipped, this opens the door to adopting Perception without starting from scratch on the hardware side.
The FlexPod reader-feeder setup shown in the demo pulling plates from storage, dispensing, and loading them into a plate reader, is just one example of how Perception gets deployed in practice (Fig. 1). The same detection and recovery logic applies across different instrument combinations and protocol types.
Figure 1. Rendering of the FlexPod with advanced Perception system, featuring the HighRes FlexPod and ACell Robotic Arm with Vision, including integrated instruments like the Liconic STX 44-SA Incubator, HighRes PreciseDrop II (PDII) Dispenser, and the Agilent Biotek Synergy Reader.
The dream of walkaway lab automation has always bumped up against the same reality: labs are messy, protocols change, humans make mistakes and the unexpected happens. Vision-based error detection and recovery is a direct answer to that gap.
What's significant about the Perception Suite isn't any single feature — it's the underlying approach. The system is designed to observe, classify, and respond to its environment dynamically. That's a meaningful shift in how automation systems relate to the unpredictability of real lab work.
Book a demo and a product specialist will reach out to you!