A COMPREHENSIVE GUIDE TO PICK AND PLACE ROBOTS AND SYSTEMS

What is the required working area and are there any spatial limitations that may impact the system’s layout and operation?

The final link in the kinematic chain of a robotic system (and anything rigidly connected to it) is referred to as the robot’s end-effector.

Each robot will have a workspace envelope specified about the robot’s origin defining the range and orientations that its end-effector can reach.

For a simple XYZ Cartesian Gantry – this envelope will be a rectangular prism defined by the travel limit of each of its axes.

A 6-DOF Articulated Arm will have a complex 6-dimensional constrained envelope describing end-effector orientation (θ_X, θ_Y, θ_Z) at different XYZ locations in space: e.g. when the arm’s ‘elbow’ joint is straightened out, its ‘wrist’ will be at the maximum radius away from its base and the end-effector orientation-range will be significantly limited compared to when the arm is bent with the end-effector closer to its base.

Is the workspace volume arbitrarily large?

If so, a single custom-scaled Cartesian Gantry may be preferable compared to requiring multiple limited-range SCARA arms working in concert.

Do the target pick and place locations exist on different work-surfaces?

Then it may be advantageous to position a SCARA robot between those surfaces rather than having a Gantry straddling them.

Is the path between each target location unimpeded or is navigation around/over interfering components required?

The exact nature of these constraints may dictate that one solution or another is preferred.

How many degrees-of-freedom at the end-effector are required by the application?

An unconstrained rigid body in space has 6 degrees-of-freedom (6-DOF): it can translate about 3-axes, D_X D_Y D_Z, and it can rotate about those same 3-axes, θ_X θ_Y θ_Z

The DOF of a robot’s end-effector, however, are constrained by the number and configuration of its joint-axes, and the relative range of motion about each of its joint-axes will dictate the absolute range of motion of the end-effector in space.

Do all pick-place target locations have the same orientation along a shared horizontal plane?

Then a simple 3DOF Gantry with minimal Z-range may be sufficient.

Do the target locations lie along multiple non-parallel planes?

(for example, if a flat component must be picked off a horizontal conveyor and re-oriented and placed against a vertical surface for processing)
Then a 6-DOF Articulated Arm is most likely required.

How will the robot interface with the world?

In a typical pick and place system, a manipulator is mounted to the end-effector that is capable of interfacing with target environment or component features.

The requirements of the manipulator should be considered when selecting a suitable pick and place system.

How large must the manipulator be?

The geometry of the manipulator must be taken into account to determine the new available working envelope for the robot’s end-effector and to evaluate interference zones.

Does the end-effector require electrical or pneumatic inputs?

Some off-the-shelf SCARA and Articulated arms provide electrical and/or pneumatic pass-throughs from their base to their end-effector to facilitate convenient connections to the manipulator, if needed. This can dramatically simplify the wire/tube routing design process compared to other systems lacking these features, (for example, a custom Cartesian Gantry), where special considerations for routing must be made such as cable/tube length, flexibility, service loops, slip-rings, etc.

How many different geometric components/features must it be able to interface with?

It is not uncommon for a single robot to have multiple manipulator tools to interface with multiple distinct component features.

This can be readily accomplished by having multiple tools rigidly mounted to the end-effector but with each having some defined offset relative from the end-effector origin.

By accounting for each tool’s offset, the robot can effectively index between each, with the consequence that each tool’s workspace volume will be offset or constrained in some manner compared to the end-effector alone.

The payload mass/inertia is the sum of those of the manipulator plus the component(s) that must be carried by the robot end-effector.

The pick and place system must be rated to accommodate the necessary payload and inertial forces.

Some pick and place system manufacturers(e.g. Epson) may spec a ‘rated’ payload in addition to a ‘max’ payload. It should be assumed that for a payload greater than the ‘rated’ spec, all other listed performance specifications (e.g. speed, acceleration, precision) will need to be derated.

A pick and place robot is often a sub-system within a larger automated process.

It is important to understand the total process throughput requirements (i.e. widgets processed per hour) to determine the maximum allowable cycle-time for the pick and place sub-system before it becomes a rate-limiting step in the overall process.

What is the minimum required accuracy/precision/repeatability required by the target application?

Although accuracy and precision (aka repeatability) are technically distinct quantities – it is common to see products specify only a single quantity (either accuracy, precision, or repeatability) defining the maximum expected position/orientation error of the end-effector across its workspace.

It is important to understand the subtle differences in interpreting this specification for different pick and place systems:

1. For a Cartesian Gantries, whose axes of motion are nominally* perpendicular and decoupled from one another – the repeatability of the end-effector about each of its component axes is simply equal to the repeatability of each corresponding drive stage – and is assumed constant across its entire workspace.
   * in reality – the drive stages are never perfectly perpendicular, resulting in some coupling of motion, and increasing error of the end-effector position from expected, with increasing distance of each stage from its home position.
2. For Serial-Articulated Robot Arms like SCARA or 6-DOF arms, the repeatability of each constituent joint axis will have some nominal constant repeatability over its range-of-motion. However, the resultant repeatability at the end-effector will vary across the robot’s workspace and is expected to be, at its worst, near the limits of the workspace volume or, more generally, when any ‘elbow’ joint is near its inversion point.

A simple rule of thumb when designing the layout of your pick and place system is to preferentially locate the target pick-place locations comfortably within the arms workspace away from its limits or elbow-inversion points.

Is the robot situated in a work environment shared by human operators?

Any automated system with actuating components introduces safety concerns for operators in their vicinity, and provisions to prevent accidental injury should be taken in their design.

Pick and place robots can be especially dangerous as they encompass multiple components that can move rapidly and forcefully across a potentially large workspace.

Typical provisions for mitigating safety concerns of automated systems are:

1. Physical isolation from human contact by being placed within an enclosure,

2. Emergency stop button present to manually deactivate the system,

3. Emergency-stop circuitry in place to automatically deactivate the system when the enclosure is opened or human presence is detected within some vicinity of the system (e.g. via a light curtain)

Does the robot have to be exposed and/or interact with human operators?

Collaborative Robots (Cobots) are specifically designed for safely interacting with or operating in a shared space alongside humans.

Cobots are typically operated with limitations imposed on their speeds and forces, and designed with either force sensors, or mechanics with software control schemes that ensure reliably safe operation in close proximity to humans.

As such, cobots are realistically the only viable option for this scenario.

Are the pick locations for incoming components in known/well-defined locations and orientations?

If the position and orientation of the incoming components are well known relative to the robot coordinate-frame – then control of the robot can effectively be run open-loop, blindly moving to defined pick locations and confidently aligning/interfacing with the components there without any additional provisions.

Are the pick locations not well defined?

If the incoming components are not in known/well-defined positions/orientations – for example, components being fed along a conveyor belt – then additional provisions are likely required to determine their position/orientation relative to the robot’s coordinate frame prior to interfacing with them, and those provisions may further constrain the selection of an appropriate pick and place system.

Determining the necessary alignment information for incoming components can be accomplished in a number of ways depending on the application.

An example approach might be to incorporate a presence detection sensor in conjunction with a vision system either mounted above the pick region or directly to the end-effector of the robot, in order to (for example):

1. Detect the presence of the next incoming component within the pick region and signal the system controller to begin the next pick cycle

2. Stop the conveyor belt

3. Identify an expected locating feature on the component (e.g. a QR code or well-defined geometric element)

4. Determine the position and orientation of that feature relative to the vision system’s coordinate frame (e.g. X, Y, Theta)

5. Feed that information to a controller and transform it into the robot’s coordinate frame

6. Send the robot to the now known position to pick the component and move it to the next step in the process

7. Restart the conveyor belt and await next incoming component

Detailed considerations for designing an appropriate vision guidance system for a pick and place application fall outside the scope of this article – but further information and available solutions can be found within Cognex’s Product Suite: Machine Vision and Industrial Barcode Reading Products | Cognex

What is the planned development schedule – how much time is allotted for design? What budgetary constraints exist for material costs, build/labor hours, non-recurring-engineering (NRE) hours for this development?

Schedule and cost are often primary driving factors in product design and development.

It is important to have an up-front understanding of any budgetary limitations which may dictate the selection of a pick and place system and the design trade-offs that can be made according to the program budget and development timeline.

What is the anticipated production quantity for the pick and place solution?
When the program plan only calls for building a small quantity of systems or if the planned development schedule is tight – it makes sense to preferentially select and integrate off-the-shelf products into the design to keep the associated NRE costs as low as possible.

For pick and place systems only requiring 3- or 4-DOF, an array of purchased off-the-shelf SCARA arms are available with varying specifications (see Epson), which, if suitable for your application, provide very cost competitive pick and place solution requiring minimal NRE to incorporate into your system.

If the planned production quantities for your system are large, however, then it might be preferable to design a bespoke Cartesian Gantry tailored to your exact requirements since the material costs will scale down with volume and the increased NRE costs associated with its design may become justifiable when spread across many builds.