Manipulation, End Effectors and Collaborative Robotics

Introduction

A robot manipulator becomes useful only when it can interact reliably with objects, tools, people, and its environment. The mechanical arm supplies positioning and orientation, but the end effector performs the physical task: grasping, inserting, welding, polishing, dispensing, fastening, or sensing. In collaborative robotics, this interaction occurs in workspaces shared with humans, so mechanical performance must be combined with force limitation, predictable behavior, ergonomic design, risk assessment, and effective task sequencing.

Manipulation is therefore a systems problem. A successful design must answer five questions:

1. What must be achieved by the task?
2. How will the object or tool be contacted and constrained?
3. Which end effector can produce the required forces and motions?
4. How much compliance and sensing are needed?
5. How can the workcell remain safe, productive, and comfortable for people?

This chapter develops these questions from contact mechanics to collaborative workcell design.

Learning Objectives

After completing this chapter, the student should be able to:

• distinguish power, precision, form-closure, and force-closure grasps;
• analyze frictional contact and estimate gripping force;
• select an appropriate end effector using technical and operational criteria;
• explain compliance, impedance, and admittance behavior;
• sequence manipulation tasks using sensing and recovery actions;
• design a human–robot collaborative workcell using risk reduction principles;
• evaluate ergonomic, ethical, and regulatory implications of collaborative robotics.

1. Manipulation and the Role of the End Effector

Manipulation is the controlled movement or transformation of an object through contact. A manipulator must control the object’s position and orientation, commonly represented by six degrees of freedom: three translations and three rotations. The final link of the robot carries the end effector, which may be an object gripper, process tool, fixture, or sensor.

An end effector has three functional layers:

• Interface: the mechanical and electrical connection to the robot wrist;
• Actuation: pneumatic, electric, hydraulic, vacuum, magnetic, or compliant mechanisms that generate motion or force;
• Task interface: fingers, pads, cups, jaws, blades, tools, or process elements that contact the object or environment.

The end effector must be compatible with payload, geometry, surface condition, cycle time, accuracy, contamination requirements, and safety constraints. A gripper that holds an object securely but damages its surface is not a successful design. Similarly, a highly accurate gripper that requires frequent manual adjustment may be unsuitable for production.

Important terminology

A grasp is the contact arrangement between an end effector and an object. A contact point is an idealized location of interaction. A contact patch has finite area and distributes pressure. Wrench denotes the combined force and moment acting on a body. Preload is the initial gripping force applied before external loads occur. Slip occurs when tangential motion exceeds the available frictional resistance. Compliance is the ability to yield under applied force or displacement.

2. Grasping Principles

2.1 Power and precision grasps

A power grasp encloses a substantial part of the object and distributes load across several fingers or the palm. It is robust against disturbance and is appropriate for heavy or irregular parts. A precision grasp uses fingertips or small contact regions to achieve dexterous positioning, but usually provides lower load capacity.

A grasp may also be classified by how it constrains motion:

• Form closure: object motion is prevented by geometry alone, ideally without relying on friction.
• Force closure: contact forces, including friction, can resist arbitrary external forces and moments within the required range.
• Friction grasp: the object is retained primarily by normal gripping forces that generate tangential friction.

In practical grippers, form and force closure often coexist. A V-groove may provide geometric location while finger preload supplies frictional retention.

2.2 Coulomb friction model

For a dry contact, the limiting tangential friction force is approximated by:

where F_t is tangential contact force in newtons, F_n is normal contact force in newtons, and μ is the dimensionless coefficient of friction. The inequality applies under a simplified Coulomb model, assuming relatively rigid surfaces, quasi-static contact, and a friction coefficient that does not vary strongly with speed, pressure, or temperature.

For a two-finger gripper lifting a body vertically, with equal normal force F_n at each finger, the ideal no-slip condition is:

Thus, the minimum normal force per finger is:

where m is object mass in kilograms and g is gravitational acceleration, approximately 9.81 m/s². In engineering design, a safety factor S is applied:

This formula neglects acceleration, vibration, moments, contact deformation, and uncertainty in μ. If the object accelerates vertically upward with acceleration a, the required force becomes:

For horizontal acceleration, the resultant inertial load should be considered. If vertical and horizontal loads are independent and the two fingers share load equally, a conservative resultant-load estimate is:

where a_v and a_h are vertical and horizontal acceleration components. This remains an idealized estimate; a detailed design must examine moments and the actual contact geometry.

Worked example: frictional gripping force

A parallel gripper lifts a 4 kg steel component. The estimated coefficient of friction between the fingers and the component is 0.40. The component experiences an upward acceleration of 1.5 m/s², and the design safety factor is 2.0. Find the minimum design normal force per finger.

The effective vertical acceleration is:

g+a = 9.81+1.5 = 11.31\ m/s²

The required force is:

F_n = (2.0×4×11.31)/(2×0.40) = 56.55\ N

Therefore, each finger should provide at least approximately 57 N of normal force under the stated assumptions. The selected gripper should provide additional capacity because friction may decrease due to oil, dust, wear, or surface variation.

2.3 Contact mechanics and pressure

The average contact pressure over a nominal contact area is:

where p_avg is pressure in pascals, F_n is normal force in newtons, and A is contact area in square metres. This average value does not describe local peak pressure at edges or asperities. Soft elastomeric pads increase effective area and friction but may deform, trap contamination, or reduce positional repeatability.

Contact models differ according to the interface. A point or line contact is a useful idealization for rigid bodies, while compliant pads create distributed contact. Hertzian contact theory can estimate local deformation for curved elastic bodies, but its assumptions—smooth surfaces, elastic deformation, and small contact regions—must be checked before use. In many gripper-selection problems, a conservative friction calculation and experimental pull or acceleration tests are more reliable than an elaborate model with uncertain material data.

Moments are important when the object’s centre of mass is offset from the grasp. If an external force F acts at perpendicular distance l from the effective grasp support, the overturning moment is:

A pair of contacts separated by distance d can resist a couple approximately proportional to the contact-force difference and separation. Increasing finger spacing, using a shaped jaw, or supporting the object from below can therefore reduce the required friction force.

3. End-Effector Selection

End-effector selection should begin with the task, not with a catalogue of available grippers. The designer should specify object mass, dimensions, centre of gravity, surface material, temperature, cleanliness, allowable marking, required orientation, cycle time, precision, and failure consequences.

A weighted decision matrix can make selection transparent. Criteria may include holding reliability, positional repeatability, cycle time, payload margin, maintenance, energy use, integration effort, safety, and total cost of ownership. A design should not optimize only purchase price: jaws, seals, sensors, utilities, downtime, and validation may dominate lifecycle cost.

Vacuum systems require attention to leakage and effective area. An ideal holding-force estimate is:

where Δp is the pressure difference between atmosphere and the vacuum side in pascals and A is the effective sealed area in square metres. Actual force is lower because of leakage, acceleration, surface roughness, cup deformation, and nonuniform loading. A vacuum gripper should therefore include vacuum sensing and a drop-detection strategy where loss of the workpiece could injure a person or damage equipment.

4. Compliance, Force Control, and Contact-Rich Tasks

A perfectly rigid robot is not always desirable. Small positioning errors can cause jamming during insertion, excessive contact force during polishing, or damage during assembly. Compliance allows the end effector or robot to absorb misalignment.

4.1 Passive and active compliance

Passive compliance is provided by mechanical elements such as springs, flexures, elastomeric joints, remote-centre-compliance devices, or floating grippers. It is intrinsically fast and can respond without computation, but its behavior is determined by mechanical stiffness and damping.

Active compliance uses force or torque sensors and robot control to regulate interaction. It can adapt to different tasks, but sensor noise, control delay, calibration error, and instability must be managed.

A simple linear spring model is:

where F is restoring force in newtons, k is stiffness in newtons per metre, and ∆x is displacement in metres. This model assumes small displacement and approximately linear elastic behavior.

An impedance controller specifies a desired dynamic relationship between displacement and force. A one-dimensional representation is:

where M_d, B_d, and K_d are desired inertia, damping, and stiffness; ∆ẋ and ∆ẍ are relative velocity and acceleration. Impedance control is useful when the robot should behave like a virtual mechanical system. Admittance control instead measures force and generates a motion response. The choice depends on robot dynamics, sensor location, bandwidth, and whether the hardware is naturally stiff or compliant.

Compliance is beneficial for insertion, deburring, surface following, and human interaction, but excessive compliance reduces positional accuracy and may permit unstable oscillation. The designer must specify a force limit, contact speed, stiffness, damping, and recovery action rather than simply making the system “soft.”

5. Task Sequencing and Manipulation Planning

A manipulation task is a sequence of states and transitions. Typical states include:

1. home or safe position;
2. approach to the object;
3. object detection and pose estimation;
4. pre-grasp alignment;
5. contact and grip closure;
6. grip verification;
7. lift and transport;
8. placement or insertion;
9. release verification;
10. return or recovery.

Each transition should have an enabling condition and a failure response. For example, the robot should not lift merely because the gripper has issued a close command. It should verify jaw position, gripping force, vacuum level, or object presence. Sensor-based sequencing is more reliable than time-based sequencing alone.

A useful task representation is a finite-state machine. At each state, the controller evaluates sensor conditions and either proceeds, retries, moves to a safe state, or requests human intervention. Recovery actions may include reopening the gripper, reorienting the object, reducing speed, presenting the part for inspection, or stopping the cell.

Planning should consider:

• collision-free robot motion and end-effector clearance;
• object stability during acceleration and deceleration;
• visibility of cameras and access for maintenance;
• handover locations and human reach;
• uncertainty in object pose and grasp position;
• sequencing of operations that change the object’s geometry;
• safe treatment of failed grasps and dropped objects.

A short cycle time is not necessarily optimal. Repeated retries, awkward human motions, or difficult recovery can reduce overall availability and increase risk.

6. Collaborative Robotics

Collaborative robotics refers to robotic systems designed for interaction with people in a shared or intermittently shared workspace. A collaborative robot is not automatically safe in every application. The complete system includes the robot, end effector, workpiece, fixtures, software, environment, and operating procedures.

6.1 Modes of collaboration

Common collaborative arrangements include:

• Safety-rated monitored stop: the robot stops when a person enters the protected space and resumes under defined conditions.
• Hand guiding: an operator directly guides the robot using an enabling device or force-sensitive interface.
• Speed and separation monitoring: robot speed or position is adjusted to maintain a validated separation distance.
• Power and force limiting: contact energy or force is limited so that specified injury risks are controlled.
• Intermittent collaboration: human and robot occupy the same area at different task stages.

These modes may be combined. The correct choice depends on hazards, contact locations, robot speed, tooling, workpiece mass, and the predictability of human movement.

6.2 Risk assessment

Risk assessment is an iterative engineering process. It should identify hazards during normal operation, teaching, setup, maintenance, fault recovery, cleaning, and foreseeable misuse. Hazards include crushing and shearing at joints, sharp tooling, ejected parts, dropped loads, unexpected restart, stored pneumatic energy, electrical hazards, ergonomic overload, and software or sensor failure.

Risk reduction follows a hierarchy:

1. eliminate the hazard through inherent design;
2. use safeguarding and protective measures;
3. provide information, training, and procedures.

Administrative instructions should not substitute for guarding or force limitation when engineering controls are feasible. Validation must test realistic worst cases, including maximum payload, failed sensors, altered friction, tool changes, and unexpected human contact.

International standards and local law must be consulted for actual deployment. Standards commonly associated with industrial robot safety include ISO 10218 and the technical specification ISO/TS 15066 for collaborative operation, while machinery risk-assessment principles are addressed by ISO 12100. Applicable national occupational-safety requirements remain authoritative. Standards should be treated as design and validation frameworks, not as permission to ignore application-specific hazards.

6.3 Ergonomic workcell design

A collaborative cell should support the human task rather than merely place a robot beside a worker. Design considerations include:

• work height and reach distance;
• neutral wrist and shoulder posture;
• low-force material presentation;
• visibility of parts, indicators, and robot motion;
• clear separation of clean and contaminated items;
• easy access for replenishment and maintenance;
• avoidance of repetitive twisting, overhead work, and sudden handover forces;
• intuitive controls and clear status indication.

The robot can present parts in a consistent orientation, perform repetitive fastening, or carry heavy components while the operator performs inspection, judgment, or delicate assembly. However, automation can also increase work pace and create psychological stress if workers cannot control the process or understand robot behavior. Ergonomic evaluation should therefore include physical, cognitive, and organizational factors.

7. Applied Case: Collaborative Bearing-Housing Assembly

The following is an illustrative case rather than a report of a particular company.

Background

A manufacturer assembles small bearing housings. An operator previously collected a housing, aligned a bearing, pressed it into place, inspected the seating condition, and transferred the assembly. The task involved repetitive reaching and occasional high insertion force.

Problem

A robot was proposed to supply housings and perform insertion while the operator completed visual inspection and installed a retaining ring. Early trials showed three problems: occasional bearing misalignment, excessive insertion force, and uncertain handover timing.

Analysis

The housing and bearing had small dimensional variation. A rigid position-controlled insertion amplified lateral error. The end effector also lacked a reliable confirmation that the bearing was fully seated. The operator had to reach across the robot’s motion path, creating a potential shared-space hazard.

Approach and solution

The cell was redesigned as follows:

• a compliant insertion unit allowed limited lateral floating and axial force control;
• a chamfered guide supported initial alignment;
• force sensing detected abnormal resistance;
• a seating-depth sensor and insertion-force signature verified completion;
• the robot presented the completed assembly at an ergonomic height;
• the operator’s inspection and retaining-ring station was placed outside the robot’s primary sweep;
• a safety-rated monitored stop was used when the operator entered the robot zone;
• failed insertions were diverted to a clearly identified inspection location;
• restart required removal of the cause and confirmation through a defined procedure.

Outcome

The redesigned sequence reduced jamming and separated normal production from fault recovery. More importantly, it made the robot’s state visible: the operator could see whether the part was awaiting inspection, accepted, or rejected. The solution did not rely on compliance alone; it combined mechanical guidance, sensing, control logic, ergonomic placement, and safeguarding.

Learning points

The case demonstrates that manipulation quality depends on the whole system. A better gripper cannot compensate for poor part presentation, inadequate sensing, or unsafe human access. Contact-rich operations should be designed around force, alignment tolerance, verification, and recovery from the beginning.

8. Contemporary Developments and Responsible Practice

Soft robotics uses compliant materials and underactuated mechanisms to grasp objects with uncertain geometry. It can reduce damage and simplify adaptation, but pressure control, durability, hygiene, and repeatability remain challenges. Vision-guided grasping and machine learning can estimate object pose or select grasps from examples. These methods are useful when parts vary, but training data may not cover unusual orientations, reflective surfaces, or damaged objects. A learned perception system should therefore be bounded by geometric checks, force limits, collision monitoring, and a safe fallback state.

Digital twins and virtual commissioning allow designers to test reachability, cycle sequences, and some collision scenarios before physical deployment. Simulation cannot fully reproduce friction, flexible objects, sensor faults, or human behavior; physical validation remains essential.

Responsible deployment also requires attention to data and employment. Cameras may capture workers, and force or productivity data may affect privacy and workplace trust. Workers should receive clear information about monitoring, participate in cell design, and be trained in normal operation and recovery. Automation decisions should consider skill development and job quality, not only throughput.

Best-Practice Design Checklist

• Define the task, object variation, and failure consequences before selecting a gripper.
• Calculate nominal holding force and apply a justified safety factor.
• Consider moments, acceleration, contamination, wear, and temperature.
• Verify grasp success using force, position, vacuum, vision, or combinations of sensors.
• Use compliance where alignment or contact uncertainty requires it.
• Limit force and speed through validated control and mechanical design.
• Model the task as states with explicit transitions and recovery actions.
• Design human reach, visibility, posture, and replenishment into the layout.
• Assess hazards in setup, maintenance, teaching, and fault recovery—not only production.
• Validate the integrated robot, tooling, workpiece, software, and safeguarding system.
• Involve operators and document residual risks, training, inspection, and maintenance.

Chapter Summary

Manipulation is the controlled interaction of a robot with objects and its environment. Grasp design depends on geometry, friction, preload, contact pressure, moments, acceleration, and uncertainty. The friction model F_t≤μF_n provides a useful first estimate, but real designs require safety factors, experimental verification, and consideration of surface and contact variation.

End-effectors should be selected from task requirements, including payload, shape, surface, accuracy, cycle time, cleanliness, maintenance, and safety. Passive and active compliance help robots tolerate misalignment and regulate contact, while impedance and admittance concepts describe force–motion behavior. Reliable manipulation uses state-based sequencing, sensor confirmation, and defined recovery actions.

Collaborative robotics is an integrated safety discipline rather than a property of a robot label. Risk assessment, safeguarding, force and speed limitation, ergonomic layout, human understanding, validation, and responsible data practices are all necessary. The most effective workcells combine robot repeatability with human judgment and design the interaction deliberately.

Key Terms

Manipulation: Controlled movement or transformation of an object through contact.

End effector: Device mounted at the robot wrist to perform a task.

Grasp: Contact arrangement used to hold or constrain an object.

Power grasp: Enclosing grasp that distributes load over several contact regions.

Precision grasp: Fingertip or localized grasp used for dexterous positioning.

Form closure: Constraint produced primarily by contact geometry.

Force closure: Ability of contact forces to resist required external forces and moments.

Preload: Initial gripping force applied before external loading.

Coefficient of friction: Dimensionless parameter relating limiting tangential and normal contact force.

Contact pressure: Normal force distributed over a contact area.

Compliance: Ability of a mechanism or controller to yield under force.

Passive compliance: Mechanical compliance provided without feedback control.

Active compliance: Compliance produced through sensing and control.

Impedance control: Control of a desired force–motion dynamic relationship.

Admittance control: Conversion of measured force into commanded motion.

Task sequencing: Ordered set of manipulation states and transitions.

Collaborative robot: Robot system designed for validated interaction or shared operation with people.

Risk assessment: Structured identification, evaluation, and reduction of hazards.

Ergonomics: Design of work and equipment to support human capability, safety, and well-being.

Review Questions

Short Answer Questions

1. Distinguish a power grasp from a precision grasp.
2. State the assumptions behind the basic Coulomb friction model.
3. What is the difference between passive and active compliance?
4. Why should a robot verify a grasp before lifting an object?
5. Name four hazards that should be considered in collaborative workcell design.

Long Answer Questions

1. Explain how object mass, friction, acceleration, contact spacing, and safety factor influence gripper selection.
2. Compare parallel-jaw, vacuum, magnetic, and adaptive grippers, including one suitable application and one limitation for each.
3. Explain impedance and admittance control and discuss when compliance is valuable in assembly.
4. Describe a state-based manipulation sequence, including sensing and recovery actions.
5. Explain why a collaborative robot is not automatically safe and describe the main elements of a complete risk assessment.

Analytical / Application Questions

1. A two-finger gripper lifts a 6 kg part with coefficient of friction 0.35 and upward acceleration 1.0 m/s². Using a safety factor of 2.5, calculate the minimum normal force per finger and state two reasons why the selected gripper should provide more capacity.
2. Design a collaborative workcell in which a robot loads parts while a worker performs inspection. Specify the task division, sensing, handover arrangement, safeguarding mode, and ergonomic features.
3. An insertion task suffers from jamming whenever part alignment varies by 1 mm. Analyze whether passive compliance, active force control, improved fixturing, or a combination is most appropriate. Justify your answer and identify validation tests.

Multiple Choice Questions with Answers

Question 1

For a two-finger friction grasp lifting a stationary mass, which ideal condition must be satisfied to prevent slip?

A. mg≤2μF_n

B. mg≤μ/F_n

C. mg≤2F_n/μ

D. mg≤F_n−μ

Correct Answer: A. mg≤2μF_n

Question 2

A 4 kg object is lifted with upward acceleration 1.5 m/s², friction coefficient 0.40, and safety factor 2.0. What design normal force per finger is required under the chapter’s two-finger model?

A. 28.3 N

B. 40.0 N

C. 56.6 N

D. 113.1 N

Correct Answer: C. 56.6 N

Question 3

Which grasp relies primarily on geometry rather than friction to constrain object motion?

A. Vacuum grasp

B. Form-closure grasp

C. Precision friction grasp

D. Magnetic preload grasp

Correct Answer: B. Form-closure grasp

Question 4

Which end effector is generally most suitable for handling ferromagnetic steel plates?

A. Magnetic gripper

B. Vacuum cup with no seal

C. Soft fingertip gripper only

D. Nonmagnetic adhesive pad

Correct Answer: A. Magnetic gripper

Question 5

In the relation F=k∆x, what does k represent?

A. Mass

B. Damping

C. Stiffness

D. Coefficient of friction

Correct Answer: C. Stiffness

Question 6

What is a principal advantage of passive compliance in insertion?

A. It eliminates all need for sensing

B. It responds mechanically without computational delay

C. It guarantees unlimited force capacity

D. It always improves positional accuracy

Correct Answer: B. It responds mechanically without computational delay

Question 7

Which action is the best example of sensor-based task sequencing?

A. Lifting after a fixed time regardless of grip condition

B. Lifting only after jaw position or force confirms grasp

C. Increasing speed whenever a fault occurs

D. Skipping release verification to reduce cycle time

Correct Answer: B. Lifting only after jaw position or force confirms grasp

Question 8

Which collaborative mode adjusts robot operation to maintain a validated distance from a person?

A. Speed and separation monitoring

B. Uncontrolled free motion

C. Manual machining mode

D. Open-loop force amplification

Correct Answer: A. Speed and separation monitoring

Question 9

Which principle belongs first in the hierarchy of risk reduction?

A. Warning labels

B. Operator memorization of hazards

C. Elimination through inherently safer design

D. Post-incident reporting

Correct Answer: C. Elimination through inherently safer design

Question 10

Why is a collaborative robot not automatically safe in every application?

A. Robot controllers cannot use sensors

B. Safety depends on the integrated robot, tool, workpiece, software, and environment

C. Collaborative robots cannot perform physical contact

D. Human operators are never allowed near robots

Correct Answer: B. Safety depends on the integrated robot, tool, workpiece, software, and environment