Teradyne connects AI perception with collaborative and mobile robots

Explore Teradyne Robotics through Universal Robots and MiR, with AI Accelerator compatibility, pallet handling and a practical buying framework. This blueprint examines documented products and proposes an evaluation; it does not report hands-on testing.

01 / ProductTwo robotics businesses put AI into different physical tasks

Teradyne supplies automated electronic test equipment as well as robotics. This article focuses on its robotics offer. Teradyne Robotics brings together Universal Robots and Mobile Industrial Robots, or MiR. Universal Robots supplies collaborative robot arms; MiR supplies autonomous mobile robots for material movement. The two businesses address different parts of factory work, so their products need separate technical and commercial evaluation.

The AI-specific offers are tangible. Universal Robots' AI Accelerator combines a Jetson Orin AGX 64GB compute module, an Orbbec Gemini 335Lg 3D camera and software for developing perception-enabled cobot applications. The product is a development platform, not a complete solution to every picking or inspection task. A buyer still needs an application and the physical tooling around it.

The MiR1200 Pallet Jack uses AI-based perception for pallet detection and is specified for EU pallets up to 1,200 kg. Here AI is embedded in a defined transport product rather than presented as a kit for creating a new application. That difference changes the buying process: one route emphasizes development flexibility, the other an accepted material-handling workflow.

02 / AudienceChoose whether the bottleneck is manipulation or movement

A machine builder developing vision-guided manipulation should begin with Universal Robots and the AI Accelerator. It may need to detect a part, determine a pose and coordinate a grasp with the existing robot program. The relevant skills include application development, camera setup, tooling and cell commissioning. Buying the kit does not remove those responsibilities.

A factory logistics team should start with the movement problem: where pallets originate, where they are needed and what causes delays. MiR may fit when repeated transport consumes attention and the route can be defined. A mobile robot cannot compensate for unclear pallet ownership, blocked destinations or a production system that does not know when material is ready.

The NVIDIA blueprint explains the wider AI compute and software environment used by many robotics developers. The Symbotic blueprint provides a useful contrast for buyers considering broader warehouse automation. Teradyne's portfolio supports component and workflow choices across robot arms and mobile transport; that breadth should not be mistaken for a single turnkey warehouse contract.

03 / WorkflowA proposed pilot should expose the handoff between perception and action

For a UR evaluation, consider a robot unloading a small set of known parts into a fixture. Start with a fixed part family and define the acceptable source presentation. The proposed test asks whether the supplied perception and development tools make that application practical. It is not a claim that Sequenced has operated the kit or measured a production pick rate.

Check the software baseline before installing anything. The current AI Accelerator documentation warns that the kit requires ROS 2 Humble and identifies PolyScope X 10.12.1 as the latest compatible release; 10.13 and later use ROS 2 Jazzy and are not supported by that documented version. A routine upgrade to the latest robot software could therefore invalidate the proposed configuration. Record the supported version set with the project.

Next, calibrate the camera and confirm the relationship between its observations and the robot's reference frames. A visually correct detection can still produce the wrong physical target when calibration or tooling offsets are wrong. Use test poses that make an error obvious before allowing a real pick. Keep the camera mounting and tool definition in the same change record as the application software.

Build the application as separate observable stages: identify the candidate object, estimate the pose, select a valid motion and confirm the result. The documentation describes PandaI in a detect, pose, plan and move sequence. For evaluation, record which stage failed. That allows the team to distinguish a perception issue from a blocked path or a gripper that cannot hold the part.

Include rejected and incomplete operations. An unknown object, an empty bin, an occluded target and a failed placement should each produce a defined response. Have an operator restore the application using the intended production interface. If recovery depends on a developer opening a console, the project may be a useful prototype while remaining unsuitable for the proposed shift pattern.

For a MiR evaluation, use a separate route study rather than assuming the cobot pilot proves transport readiness. Choose a pallet flow with clear pickup and delivery conditions. Record pallet dimensions, wrapping, load stability, aisle constraints and shared traffic. Test the complete mission, including waiting for an occupied destination, returning to charge and recovering from an unavailable pickup.

Assess throughput as completed deliveries over the actual operating window. Maximum travel speed is not the same as sustained transport capacity. Waiting, loading, charging and traffic determine the useful output. Compare the mobile route with the current process at the same demand profile, and keep peak congestion visible instead of averaging it away.

Only connect the two pilots when each has an accepted handoff. A robot arm may place a completed load in a defined pickup area, while the mobile workflow confirms that the destination can receive it. Define who owns the material during each transition. That operational agreement is more valuable than showing two robots moving in the same demonstration.

04 / PricingOrderable hardware does not establish a complete application price

The AI Accelerator product page says the kit is available to order and lists bundles with the accelerator alone, a control box, or a control box and e-Series robot. The inspected page does not expose a universal currency price. A quote needs to identify the bundle, existing robot compatibility and any additional hardware needed for the application.

MiR's product page similarly supports a product inquiry rather than a public fixed price for an installed pallet workflow. The Teradyne contact directory provides routes to the relevant business. Keep hardware, application integration, fleet operation and service obligations separate in the proposed scope so the buyer can compare equivalent deliverables.

A useful cost model includes engineering changes after launch. A new part may require camera or model work; a new pallet route may need traffic and destination changes. Ask which changes the local team can perform, which require the supplier and how support is priced. Avoid substituting a promotional payback claim for a calculation based on the site's actual workload.

05 / DistinctionsThe portfolio offers both a development platform and a defined product

The collaborative robot overview explains UR's role within Teradyne, while the mobile robot overview establishes MiR's material-handling role. Their common ownership is useful context, but an end user should still expect different product interfaces, implementation skills and acceptance criteria.

The AI Accelerator exposes a route for building differentiated applications on a robot platform. The MiR1200 packages perception around a narrower task. Neither approach is inherently better. A developer may value access to software components; a logistics manager may value a constrained product whose behavior can be accepted against a defined route.

The strongest reason to consider Teradyne is therefore the fit of a specific product to the work. Our editorial assessment is that combining arm and mobile robotics can become useful after the process boundaries are clear. Starting with a large coordinated demonstration before establishing those boundaries risks spending engineering effort on orchestration while the individual tasks remain unreliable.

06 / QuestionsVersion compatibility and real pallets deserve early attention

For the accelerator, obtain the currently supported software matrix in writing and include it in maintenance planning. The explicit documentation warning is more actionable than a general statement that PolyScope X is supported. Ask how upgrades, security maintenance and application dependencies will be coordinated over the installation's life.

For MiR1200, verify the actual pallet types, load geometry and floor conditions against the supplier's supported application. A quoted payload limit does not establish every load's stability or suitability. Include the site's ordinary damaged, wrapped or misplaced pallets in the evaluation discussion and agree which should be rejected before a mission begins.

For both routes, define the responsibility for restoring service. The physical equipment, application software and production system may be supported by different parties. A useful support arrangement identifies who investigates a failed operation and what evidence is available to them. That can matter more to production continuity than a fast initial demonstration.

07 / DecisionBuy the capability at the layer where the problem occurs

Teradyne offers a substantial AI-related robotics portfolio through UR and MiR. Choose the accelerator when the work calls for application development on a cobot; choose the pallet-handling route when the problem is a defined internal transport flow. In either case, accept the complete workflow under real operating conditions before expanding the deployment.