11. Mobile, Autonomous and Collaborative Robotics

Introduction and Context

Industrial robotics was historically associated with fixed-base manipulators operating inside carefully controlled work cells. That model remains valuable for repetitive manufacturing, but many contemporary operations require robots to move through changing environments, interact with people, handle variable objects, and make decisions with incomplete information. Warehouses, hospitals, farms, construction sites, mines, disaster zones, and service facilities therefore demand a broader class of robotic systems: mobile robots, autonomous mobile robots (AMRs), field robots, and collaborative robots (cobots).

This chapter examines the integration of mobility, perception, estimation, planning, manipulation, and human interaction. A mobile robot must answer three interdependent questions: “Where am I?”, “What exists around me?”, and “How should I act?” Localization and mapping address the first two questions; navigation and control address the third. When manipulation is added, the robot must also coordinate the motion of a mobile platform with that of an arm. In collaborative applications, technical capability must be combined with risk assessment, human factors, and governance.

The subject is consequently not merely an extension of industrial automation. It is a systems-engineering problem involving mechanics, sensors, embedded computation, artificial intelligence, control, communications, safety engineering, and operations management. The quality of a deployed system depends less on any single algorithm than on how reliably the complete architecture performs under uncertainty.

Learning Objectives

After studying this chapter, the reader should be able to:

1. Distinguish major mobile robot locomotion systems and select an appropriate configuration for a task.

2. Explain coordinate frames, motion models, localization, mapping, and simultaneous localization and mapping (SLAM).

3. Compare navigation architectures, path-planning methods, and obstacle-avoidance strategies.

4. Analyse autonomous manipulation and mobile-manipulator coordination.

5. Explain the operation, capabilities, and limitations of AMRs, cobots, and field robots.

6. Evaluate fleet-coordination approaches and the role of digital infrastructure.

7. Identify safety, ethical, cybersecurity, regulatory, and organizational issues in collaborative autonomy.

8. Apply suitable design and validation practices to a realistic robotic deployment.

From Locomotion to Autonomy

Locomotion and mechanical architecture

The mechanical design establishes the set of motions that a robot can produce and the environments in which it can operate. Wheeled robots are efficient on prepared floors. Differential-drive platforms use two independently driven wheels and one or more passive supports; they are inexpensive and highly manoeuvrable, but cannot move sideways instantaneously. This is a nonholonomic constraint: the vehicle’s lateral velocity is not independently controllable.

Ackermann-steered vehicles, resembling automobiles, are appropriate for outdoor roads and larger platforms. Their steering geometry reduces tyre scrub but imposes a larger turning radius. Omnidirectional platforms, often using mecanum or omni-wheels, can translate laterally and rotate in place. Their flexibility is useful in congested factories, although wheel-ground contact, payload, floor irregularities, and control calibration affect performance.

Tracked vehicles distribute contact over a larger area and can traverse loose terrain, but turning may cause substantial slip and energy loss. Legged robots can negotiate steps, rocks, and discontinuous terrain, yet require complex balance control, consume considerable energy, and generally have lower payload efficiency. Aerial and aquatic robots extend mobility into three-dimensional environments, but their endurance, communication, and safety constraints differ greatly from those of ground robots.

Locomotion should be selected from the mission rather than from novelty. Relevant criteria include terrain, floor quality, slope, payload, speed, turning space, energy availability, noise, maintenance, human proximity, and recovery from wheel slip or loss of contact.

Mathematical description of movement

A mobile robot is commonly described using a world frame, a robot or body frame, and sensor frames. A planar pose is represented as

q = [x, y, theta]^T,

where x and y describe position and theta describes heading. For a differential-drive robot with wheel radius r, wheel separation L, and right and left wheel angular velocities omega_R and omega_L, the ideal body velocities are

v = r(omega_R + omega_L)/2,

omega = r(omega_R - omega_L)/L.

The continuous kinematic model is

dot{x} = v cos(theta), dot{y} = v sin(theta), dot{theta} = omega.

This model is useful for planning and control, but real robots experience wheel slip, backlash, unequal wheel radii, floor irregularity, and timing errors. Consequently, odometry provides an estimate rather than an exact pose. Calibration and sensor fusion are essential, particularly when a robot must operate for long periods without external references.

The autonomy stack

A practical autonomous robot is usually organized into layers. The lowest layer contains actuators, motor controllers, emergency stops, and real-time safety functions. The estimation layer combines wheel encoders, inertial measurement units (IMUs), cameras, lidar, radar, ultrasonic sensors, or satellite positioning. The perception layer detects obstacles, surfaces, objects, people, and semantic features. A world-model layer maintains maps and tracked entities. Planning generates a route and a feasible local motion. Control converts that motion into actuator commands. A mission or task layer assigns goals such as transporting a container or inspecting a pipeline.

This layered arrangement is not absolute. Modern systems may use behaviour trees, state machines, optimization-based controllers, or learning-based policies that cross traditional boundaries. Nevertheless, separating safety-critical functions from experimental intelligence remains sound engineering practice. A learned perception model may propose that an object is a pallet; a certified controller must still ensure that the vehicle stops if a person enters the protective zone.

Perception, Localization and Mapping

Sensors and uncertainty

No sensor provides a complete description of the environment. Wheel encoders provide high-rate motion information but accumulate drift. IMUs respond quickly but suffer from bias and integration error. Lidar gives accurate geometric ranges and is relatively insensitive to illumination, while cameras provide rich appearance and semantic information but are affected by lighting, texture, occlusion, and motion blur. Radar can be valuable in dust, rain, or fog, though its spatial resolution may be lower. Global navigation satellite systems support outdoor localization but can be unavailable or unreliable indoors, underground, or near structures.

Sensor fusion combines complementary measurements. In a probabilistic formulation, the robot maintains a belief about its state rather than a single unquestionable value. The Kalman filter and its extensions are effective when models and noise assumptions are approximately known. Particle filters represent non-Gaussian hypotheses and are widely used for global localization. The central principle is that predictions from a motion model are corrected by observations, with the relative confidence determined by uncertainty.

Localization and mapping

Localization estimates the robot pose relative to a known map. Mapping estimates environmental structure from sensor observations. If both the pose and map are unknown, the problem becomes simultaneous localization and mapping. SLAM is difficult because an error in pose changes the inferred position of every observed feature, while an incorrect map also corrupts future localization.

A SLAM system typically includes front-end data association and motion estimation, a back-end optimization or filtering process, and loop-closure detection. The front end identifies correspondences between successive sensor observations. The back end adjusts a trajectory and map to reduce measurement inconsistency. A loop closure occurs when the robot recognizes a previously visited location; this provides a constraint that can correct accumulated drift. Pose-graph optimization represents robot poses as nodes and relative measurements as edges. Graph optimization then seeks a globally consistent configuration.

Maps may be metric, topological, or semantic. An occupancy grid divides space into cells with estimated probabilities of being occupied. A topological map represents places and connections, often at lower computational cost. A semantic map adds meanings such as “charging station,” “fire door,” or “pallet storage area.” Autonomous systems increasingly use hybrid maps because geometric precision alone does not specify what actions are permitted.

Localization failure may arise from repetitive corridors, dynamic obstacles, poor lighting, feature-poor surfaces, map changes, or sensor degradation. Recovery strategies include relocalization against a global map, use of additional landmarks, multiple hypotheses, human assistance, and safe stopping. A robust design must define what the robot does when confidence falls below an acceptable threshold.

Navigation: From a Goal to Safe Motion

Navigation is often divided into global planning, local planning, and control. Global planning uses a relatively stable map to find a route from the current position to a goal. Dijkstra’s algorithm guarantees a shortest path on a non-negative graph, while A* improves efficiency by using a heuristic estimate of remaining cost. The cost need not be distance alone; it may include time, energy, congestion, slope, risk, or restricted zones.

The resulting path is not automatically executable. A local planner considers the robot’s dimensions, velocity limits, acceleration, steering constraints, and moving obstacles. It may generate and score short trajectories, use the dynamic window approach, apply model predictive control, or optimize a trajectory directly. Obstacle inflation is commonly used in grid planning: occupied cells are expanded by a safety margin based on robot geometry and uncertainty.

A navigation controller minimizes errors between the desired and actual trajectory. Proportional-integral-derivative control may be adequate for simple platforms, while model predictive control is useful when constraints and predictions must be considered explicitly. Planning and control must be coordinated. A path that is geometrically short but requires an infeasible curvature is inferior to a slightly longer path that respects the vehicle’s dynamics.

Dynamic environments require prediction and social awareness. A human walking across a corridor is not simply a moving obstacle; the robot should avoid surprising or intimidating behaviour. Socially acceptable navigation may require personal-space models, visibility, passing conventions, speed reduction, and communication through lights, sounds, or displays. These behaviours are context-dependent and should be validated with representative users rather than assumed from mathematical distance alone.

Autonomous Manipulation and Mobile Manipulators

A mobile manipulator combines a mobile base with an arm, end effector, and perception system. The base increases the reachable workspace, but it also introduces additional degrees of freedom and uncertainty. Manipulation begins with object detection and pose estimation, followed by grasp selection, collision checking, arm and base planning, grasp execution, and force or visual feedback.

A grasp must satisfy geometric and physical requirements. The grasp configuration should provide sufficient contact forces and moments to resist gravity and task loads, while accommodating uncertainty in object pose, friction, and compliance. Suction, parallel-jaw, underactuated, magnetic, and soft grippers each suit different materials and shapes. End-effector selection is often more consequential than the choice of planning algorithm.

The base and arm can be planned sequentially or jointly. In sequential planning, the base first moves to a manipulation position and the arm then performs the task. This is computationally simpler but may fail when the arm cannot reach or when the base must adjust during contact. Whole-body planning treats base and arm configuration together, allowing coordinated repositioning. Visual servoing uses image features to guide motion, whereas force control regulates contact forces during insertion, polishing, or compliant assembly.

Uncertainty is especially significant in manipulation. A robot may detect a box accurately but still fail because the box is flexible, partly occluded, slippery, or heavier than expected. Robust systems use regrasping, tactile feedback, compliance, grasp verification, and recovery behaviours rather than treating the first plan as final.

AMRs, Cobots and Field Robots

An autonomous mobile robot differs from a conventional automated guided vehicle primarily in the degree of environmental adaptability. Traditional guided vehicles often follow fixed wires, magnetic markers, reflectors, or predefined routes. AMRs generally use maps, onboard perception, and dynamic planning to select routes and avoid obstacles. This flexibility reduces dependence on fixed infrastructure, but it shifts responsibility toward software, sensing, map management, and operational governance.

Cobots are designed to work in proximity to people, often assisting with assembly, machine tending, inspection, packaging, or ergonomic lifting. Their collaborative capability is not guaranteed merely by low mass or force sensing. It depends on the complete application, including tool geometry, payload, speed, reach, workpiece hazards, workspace layout, and foreseeable human behaviour. A cobot carrying a sharp tool can remain hazardous even if the arm itself has compliant joints.

Field robots operate in environments that are less structured and less predictable than factories. Agricultural robots may identify crops, weeds, or fruit while coping with changing weather and terrain. Inspection robots may enter tunnels, pipelines, substations, or offshore facilities. Search-and-rescue robots must tolerate damaged infrastructure, degraded communications, smoke, dust, and uncertain maps. For these systems, autonomy is often supervisory rather than absolute: a human operator sets objectives, reviews exceptions, and intervenes when confidence is low.

Fleet Coordination and Digital Infrastructure

A single robot may be technically successful yet operationally ineffective if it conflicts with other robots, blocks shared resources, or arrives at a workstation when no operator is available. Fleet management assigns missions, allocates chargers, coordinates traffic, monitors health, and records performance. Centralized systems simplify global optimization and data management but create communication and single-point-of-failure concerns. Distributed approaches can improve resilience and scalability, though coordination becomes more complex.

Common coordination mechanisms include traffic graphs, reservation of narrow passages, time-window planning, priority rules, task auctions, and multi-agent path finding. Fleet objectives can be expressed through makespan, throughput, travel distance, battery consumption, lateness, and congestion. A useful objective is not necessarily the shortest individual route; it is the best system-level performance under operational constraints.

Interoperability is a major concern. Robots, warehouse-management systems, manufacturing-execution systems, elevators, doors, and charging infrastructure must exchange reliable information. Standardized interfaces and explicit semantic models reduce dependence on a single vendor. Digital twins can support simulation, what-if analysis, predictive maintenance, and commissioning, but their value depends on the accuracy and timeliness of the data they represent.

Human–Robot Collaboration and Safety

Human–robot collaboration requires a risk-based design process. Typical protective measures include safety-rated monitored stop, hand guiding, speed and separation monitoring, and power-and-force limiting. These functions are not substitutes for risk assessment. A deployment team must identify hazards from robot motion, payloads, pinch points, dropped objects, unexpected restart, software faults, and interaction with nearby equipment.

Safety has both physical and functional dimensions. Physical safety limits contact energy and prevents access to hazardous motion. Functional safety ensures that protective sensors, controllers, brakes, and emergency-stop circuits perform reliably when demanded. Safety-related control systems require validation, fault analysis, defined performance levels, and controlled changes. Applicable international and national standards should be consulted for the specific robot and application; compliance is an engineering obligation, not a marketing label.

Human factors are equally important. Workers need clear indications of robot intent, understandable operating modes, accessible stop controls, training, and a realistic opportunity to report unsafe behaviour. Excessive false alarms can cause alarm fatigue, while opaque autonomy can encourage inappropriate trust. The design should communicate uncertainty and define who has authority during conflicts between production and safety.

Ethical and governance issues arise when robots collect video, location, biometric, or productivity data. Data minimization, access control, retention policies, and transparent purpose limitation are necessary. Autonomous decisions affecting workers should be auditable. Organizations should also consider job redesign, skills development, responsibility for failures, accessibility, and the possibility that automation transfers rather than removes risk.

Artificial Intelligence and Emerging Technical Directions

Machine learning is increasingly used for object recognition, semantic segmentation, anomaly detection, grasp selection, predictive maintenance, and behaviour prediction. Reinforcement learning and imitation learning can produce effective policies in simulation or constrained environments, but deployment requires attention to distribution shift, rare events, interpretability, and safety guarantees. Simulation-to-reality transfer is improved through domain randomization, system identification, and real-world validation.

Foundation models and natural-language interfaces may make robots easier to instruct, but language-level reasoning does not remove the need for geometric verification and safety interlocks. A system that interprets “move the container” must identify the correct container, verify its mass and graspability, and ensure that the route is safe. Edge computing reduces dependence on cloud connectivity and limits the exposure of sensitive data, whereas cloud systems support fleet-level learning and large-scale analytics.

Cybersecurity is now part of robot safety. Threats include unauthorized mission changes, manipulated maps, denial of service, stolen credentials, and compromised software updates. Secure boot, network segmentation, authentication, signed updates, logging, vulnerability management, and recovery procedures should be incorporated from the design stage.

An Applied Case: AMR-Based Intralogistics with Human Picking

Background

The following is an illustrative case. A medium-sized manufacturing plant supplies several assembly lines from a central supermarket. Previously, forklifts and manual carts delivered kits. Congestion, inconsistent delivery times, and near misses at pedestrian crossings reduced productivity.

Situation and problem

Management proposed a fleet of AMRs to transport standardized bins. The plant contained narrow aisles, mixed pedestrian traffic, changing inventory locations, and two elevators. A simple shortest-path system was rejected because it could create queues at crossings and would not coordinate elevator access.

Analysis

A site survey identified floor irregularities, reflective surfaces, blind corners, fire exits, charging locations, and areas in which workers frequently stopped to discuss tasks. Lidar and cameras were selected for complementary geometric and semantic perception; wheel odometry and inertial sensing supported short-term estimation. The map was divided into travel lanes, safety zones, staging areas, and restricted regions. Historical delivery data indicated that throughput depended more on reliable task release and congestion control than on maximum robot speed.

Approach and solution

The deployment team first introduced one robot during off-shift trials. A pose-graph map was created, then verified against physical measurements and emergency routes. Global planning used a cost map that penalized congestion and proximity to high-traffic pedestrian zones. A local planner enforced vehicle dimensions, stopping distance, and speed limits. The fleet manager reserved elevator time windows and prevented excessive robot accumulation at a workstation.

Human interaction was addressed through visible status indicators, audible but non-startling alerts, marked pedestrian crossings, and a policy that robots yield at designated conflict points. Workers received training on robot intent, manual recovery, and emergency stopping. The system logged localization confidence, blocked-path events, intervention frequency, battery state, delivery lateness, and near misses.

Rather than allowing autonomous recovery in every circumstance, the system entered a safe stop when localization confidence fell below a threshold, when a sensor became unavailable, or when a route was persistently blocked. A supervisor could approve rerouting after inspecting the event. Cybersecurity measures included role-based access, segmented wireless networks, authenticated software updates, and audit logs.

Outcome and learning points

The pilot demonstrated that technical navigation performance alone was not a sufficient acceptance criterion. Reliable operation depended on map governance, traffic rules, workstation readiness, worker trust, and clear exception handling. The most important learning points were:

• Introduce autonomy incrementally and validate it in the real operating context.
• Design the fleet and facility together; robot capability cannot compensate for poor traffic layout.
• Measure system outcomes such as delivery reliability and interventions, not only path accuracy.
• Treat uncertainty and degraded modes as normal operating conditions.
• Involve workers in hazard identification and workflow redesign.

Limitations, Failure Modes and Best Practices

Typical failures include assuming that a static map remains valid, confusing detection with understanding, ignoring wheel slip, using a planner that violates vehicle dynamics, and testing only in ideal lighting or empty aisles. Manipulation systems often fail because grasp uncertainty and object variability were underestimated. Fleet projects fail when charging, elevators, doors, and enterprise software are treated as secondary details.

A disciplined development process includes requirements analysis, operational design domain definition, hazard analysis, sensor and actuator selection, simulation, component testing, subsystem integration, staged pilot deployment, performance monitoring, and periodic reassessment. Key metrics may include localization error, route completion rate, minimum separation, intervention frequency, energy per task, task cycle time, collision and near-miss reports, and system availability. Metrics should be interpreted together: a low intervention rate is not positive if operators have stopped reporting problems.

Best practice is to define graceful degradation. When perception is uncertain, the robot should slow or stop; when communication is lost, it should follow a known safe policy; when a mission cannot be completed, it should report the reason and preserve the state needed for recovery. Human override should be deliberate, logged, and safe. Finally, every deployment should have an ownership model covering maps, software versions, safety validation, data, cybersecurity, maintenance, and incident response.

Current Directions

Research and industry are moving toward more capable multimodal perception, shared autonomy, soft and adaptive grippers, legged and hybrid mobility, energy-aware fleet scheduling, and learning systems that improve from operational data. Mobile manipulation is progressing from fixed pick-and-place demonstrations toward open-world tasks, although reliability remains a central challenge. Collaborative robotics is also expanding from physical proximity to cognitive collaboration, in which robots explain plans, request assistance, and adapt to operator expertise.

The important engineering trend is convergence. Mobility, manipulation, artificial intelligence, safety, and enterprise integration are increasingly designed as one cyber-physical system. Future progress will therefore depend not only on better algorithms, but also on verifiable behaviour, interoperable platforms, maintainable architectures, and responsible organizational adoption.

Chapter Summary

Mobile and autonomous robotics combines locomotion, sensing, estimation, mapping, planning, control, and task execution under uncertainty. Wheeled, tracked, legged, and other locomotion systems present different trade-offs in efficiency, terrain capability, payload, and control complexity. Localization estimates pose, mapping represents the environment, and SLAM jointly estimates both while correcting drift through loop closure and optimization.

Autonomous navigation normally combines global planning, local trajectory generation, and feedback control. Mobile manipulation extends the problem by coordinating a moving base, arm, end effector, perception, grasping, and contact control. AMRs provide adaptable material movement, cobots support human-adjacent work when the complete application is risk-assessed, and field robots operate in unstructured environments with greater uncertainty.

Fleet coordination, digital integration, human factors, cybersecurity, and governance determine whether a technically capable robot creates operational value. Safe and effective deployment requires uncertainty handling, staged validation, clear degraded modes, worker participation, performance metrics, and accountable ownership. Emerging AI capabilities are important, but they must be integrated with geometric verification, safety engineering, and responsible data practices.

Key Terms

SLAM: Simultaneous estimation of a robot’s pose and an unknown map.

AMR: Autonomous mobile robot that uses onboard perception and planning to move through changing environments.

Cobot: Robot designed for applications involving close human proximity, subject to application-specific risk assessment.

Nonholonomic constraint: A motion constraint in which certain instantaneous velocities cannot be independently commanded.

Odometry: Motion estimation obtained from wheel or joint displacement measurements.

Occupancy grid: Spatial map whose cells contain estimated probabilities of being occupied.

Loop closure: Recognition of a previously visited place used to correct accumulated localization drift.

Pose graph: Graph of robot poses and spatial constraints used for trajectory optimization.

Sensor fusion: Combination of multiple sensor measurements to obtain a more reliable state estimate.

Global planner: Planner that computes a route over a relatively stable map.

Local planner: Planner that generates feasible short-horizon motion around current obstacles and constraints.

Model predictive control: Control method that repeatedly optimizes predicted future behaviour subject to constraints.

Mobile manipulator: Robotic system combining a mobile platform with an arm and end effector.

Visual servoing: Robot control using visual measurements or image features as feedback.

Fleet management: Coordination of missions, traffic, charging, health, and resources among multiple robots.

Graceful degradation: Safe reduction of capability when sensing, communication, or computation becomes unreliable.

Collaborative operation: Robot operation in which people and robots share a workspace or task context.

Operational design domain: Defined set of environmental and operational conditions in which an autonomous system is intended to function.

Digital twin: Data-linked virtual representation of a physical asset, process, or facility.

Review Questions

Short Answer Questions

1. Why is a differential-drive robot described as nonholonomic?

2. What problem does loop closure solve in SLAM?

3. Distinguish global planning from local planning.

4. Why does a cobot require application-specific risk assessment?

5. State two functions commonly performed by a fleet manager.

Long Answer Questions

1. Explain the relationship among sensing, localization, mapping, planning, and control in an autonomous mobile robot.

2. Compare wheeled, tracked, and legged locomotion for industrial and field applications.

3. Describe the principal stages of autonomous manipulation by a mobile manipulator.

4. Discuss the safety and human-factors requirements of a collaborative robotic work cell.

5. Explain how uncertainty, degraded modes, and cybersecurity should influence AMR deployment.

Analytical / Application Questions

1. Select a locomotion system for a robot that must transport loads across smooth factory floors and occasionally enter narrow aisles. Justify the choice using mechanical and operational criteria.

2. An AMR repeatedly loses localization in a visually repetitive corridor containing moving workers. Diagnose possible causes and propose a sensor, mapping, and operational strategy.

3. Design a validation plan for the illustrative AMR fleet case. Specify performance metrics, safety tests, human-involvement activities, and criteria for expanding beyond the pilot.

22. Multiple Choice Questions with Answers

Question 1

Which statement best explains why odometry alone is insufficient for long-duration mobile-robot navigation?

A. Odometry cannot measure wheel rotation.

B. Odometry accumulates errors from slip, calibration errors, and mechanical imperfections.

C. Odometry operates only when satellite positioning is available.

D. Odometry directly measures semantic object categories.

Correct Answer: B. Odometry accumulates errors from slip, calibration errors, and mechanical imperfections.

Question 2

What is the principal function of loop closure in a SLAM system?

A. To increase motor torque during turning.

B. To identify a previously visited place and reduce accumulated map and trajectory drift.

C. To replace all sensor-fusion algorithms.

D. To convert a metric map into a manufacturing schedule.

Correct Answer: B. To identify a previously visited place and reduce accumulated map and trajectory drift.

Question 3

A global planner produces a geometrically short route that requires curvature greater than the vehicle can achieve. What is the most appropriate response?

A. Execute the route at a higher speed.

B. Ignore the curvature because global planners do not use maps.

C. Use a feasible local planner or replanning method that respects vehicle dynamics.

D. Disable obstacle inflation.

Correct Answer: C. Use a feasible local planner or replanning method that respects vehicle dynamics.

Question 4

Which feature most clearly distinguishes an AMR from a vehicle following a fixed guide wire?

A. The AMR always has a larger payload.

B. The AMR uses perception and planning to adapt routes to environmental changes.

C. The AMR cannot operate indoors.

D. The AMR does not require sensors.

Correct Answer: B. The AMR uses perception and planning to adapt routes to environmental changes.

Question 5

Why must a mobile manipulator coordinate base and arm motion?

A. The arm has no actuators of its own.

B. Base positioning affects reachability, collision avoidance, and manipulation accuracy.

C. The base determines the colour of the object being grasped.

D. Coordination eliminates the need for object detection.

Correct Answer: B. Base positioning affects reachability, collision avoidance, and manipulation accuracy.

Question 6

Which situation is an example of graceful degradation?

A. The robot continues at full speed after losing a critical range sensor.

B. The robot silently deletes its map after a communication failure.

C. The robot slows or enters a safe stop when localization confidence falls below a defined threshold.

D. The robot bypasses its emergency-stop circuit to complete a mission.

Correct Answer: C. The robot slows or enters a safe stop when localization confidence falls below a defined threshold.

Question 7

Why is a cobot’s low arm force not, by itself, proof that an application is safe?

A. Collaborative safety depends on the complete application, including tools, payloads, speed, and foreseeable contact hazards.

B. Low force always indicates a software failure.

C. Cobots are safe only when used outdoors.

D. Arm force has no relationship to risk.

Correct Answer: A. Collaborative safety depends on the complete application, including tools, payloads, speed, and foreseeable contact hazards.

Question 8

In fleet coordination, why might the shortest individual route fail to optimize system performance?

A. Individual routes never affect other robots.

B. Short routes can increase congestion, queuing, energy use, or lateness elsewhere in the fleet.

C. Fleet managers are prohibited from using maps.

D. Shortest paths are always physically impossible.

Correct Answer: B. Short routes can increase congestion, queuing, energy use, or lateness elsewhere in the fleet.

Question 9

Which map type adds meanings such as “charging station” or “restricted region” to geometric information?

A. Semantic map

B. Raw odometry log

C. Motor-current map

D. Purely topological-free map

Correct Answer: A. Semantic map

Question 10

A robot deployment records low collision counts but frequent operator interventions and unreported near misses. What is the most appropriate interpretation?

A. The system is automatically successful because collisions are rare.

B. The metrics are incomplete and the deployment requires investigation of interventions, reporting culture, and near-miss data.

C. The robot should be allowed to move faster.

D. Human involvement should be removed immediately.

Correct Answer: B. The metrics are incomplete and the deployment requires investigation of interventions, reporting culture, and near-miss data.