Physical Human-Robot Interaction (pHRI)

Compliant actuators are joint motors designed to be backdrivable and force-sensitive, allowing them to yield to external forces rather than rigidly opposing them. This is essential for safe physical contact. Key types include:

• Series Elastic Actuators (SEAs): Use a physical spring between the motor and output to measure force and absorb impacts.
• Variable Stiffness Actuators (VSAs): Can dynamically adjust their mechanical stiffness to suit different tasks, from precise positioning to shock absorption.
• Torque-Controlled Motors: Directly measure and control output torque with high fidelity, enabling sensitive force feedback. These components form the physical foundation for implementing control strategies like impedance and admittance control.

These are the two primary force control paradigms for pHRI, defining how a robot responds to physical interaction.

• Impedance Control: The robot regulates its dynamic relationship between position and force. It acts like a programmable spring-damper system: when pushed, it yields with a defined stiffness and damping. The controller commands torque based on position/velocity error.
• Admittance Control: The robot regulates its motion in response to measured forces. It acts like a programmable mass: when a force is sensed, it moves with a defined inertia and damping. The controller calculates a desired acceleration/velocity from force input. Impedance control is often implemented at the joint level, while admittance control is common in industrial cobots where high-geared, non-backdrivable motors are used with a six-axis force/torque sensor at the wrist.

Accurate measurement of interaction forces is non-negotiable for pHRI. This is achieved through specialized sensors:

• Six-Axis Force/Torque (F/T) Sensors: Typically mounted at the robot's wrist or end-effector, these strain-gauge-based sensors measure the three force components (Fx, Fy, Fz) and three torque components (Tx, Ty, Tz) applied at that point. They are critical for admittance control, contact detection, and delicate manipulation tasks.
• Joint Torque Sensors: Integrated into the actuator to directly measure the output torque of each joint. This provides a more direct signal for impedance control and collision detection across the entire arm.
• Tactile/Skin Sensors: Distributed arrays of pressure sensors on the robot's surface to detect the location and magnitude of contact, enabling whole-body safety and complex manipulation.

A real-time safety system that identifies unintended contact and triggers a protective stop. It operates through two main methods:

• Model-Based Monitoring: Compares the expected joint currents/torques (from the dynamic model of the robot) with the actual measured values. A significant discrepancy indicates a collision. This works without external sensors.
• Direct Force Sensing: Uses data from F/T sensors or tactile skin to detect contact forces exceeding a safe threshold. Upon detection, the robot must react within milliseconds. Standard reactions defined in ISO/TS 15066 include:
• Safety-rated monitored stop: Immediate cessation of all motion.
• Joint-level torque limiting: Actively resisting only up to a safe force level.
• Retraction: Moving away from the contact point along a safe trajectory.

These are the designed points of contact where force and information are exchanged between human and robot.

• Hand-Guiding Handles: Ergonomic handles with integrated buttons and force/torque sensors. The human can physically lead the robot (kinesthetic teaching), with the robot sensing the intended direction and magnitude of force.
• Compliant End-Effectors: Grippers or tools with inherent mechanical compliance or soft materials (e.g., silicone, pneumatic actuators) that conform to objects and safely distribute contact forces.
• Wearable Interfaces: Exoskeletons or haptic devices that allow the human to feel the robot's forces or vice-versa, enabling bilateral teleoperation for high-precision remote tasks. The design prioritizes intuitive force application, clear affordances, and passive safety through form factor.

pHRI systems must be validated against rigorous technical standards. The cornerstone is ISO/TS 15066, which supplements the broader machinery safety standard ISO 10218. It provides:

• Quantitative limits for Power and Force Limiting (PFL), including permissible maximum forces and pressures for different body regions (e.g., hand, face).
• Guidelines for the four collaborative operation modes: Safety-rated monitored stop, Hand guiding, Speed and separation monitoring, and Power and force limiting.
• Methodology for collaborative application risk assessment, requiring analysis of all foreseeable contact scenarios. Compliance involves a combination of inherent design (rounded edges, compliant covers), control functionality (force/torque limits), and performance validation through physical testing with biofidelic measurement devices.

Power and Force Limiting (PFL) is a collaborative operational mode where the robot is designed or controlled to inherently restrict the power and force of its movements to levels considered safe for incidental or expected contact. Implementation strategies include:

• Inherent Design: Using lightweight materials, rounded edges, and compliant joints or covers.
• Active Control: Employing impedance control or admittance control to limit joint torques and end-effector forces.
• Application-Specific Risk Assessment: Engineers must calculate maximum possible forces based on robot mass, velocity, and geometry, ensuring they stay below ISO/TS 15066 limits for the relevant body region.

Speed and Separation Monitoring (SSM) is a safety mode that maintains a protective separation distance between the robot and a human. The system dynamically adjusts the robot's speed based on real-time tracking of the human's position and velocity. Core components are:

• Protective Separation Distance (S_p): A calculated zone that must never be breached. It factors in robot stopping time, human intrusion speed, and system latency: S_p = (v_r * T_r) + (v_h * (T_r + T_s)) + C + Z_d + Z_r.
• 3D Sensing Array: Uses LiDAR, depth cameras, or safety-rated laser scanners to monitor the shared workspace.
• Hierarchical Speed Reduction: The robot slows as the human approaches and executes a protective stop before contact is possible.

These are two additional collaborative modes defined by ISO/TS 15066 for specific interaction scenarios.

Safety-Rated Monitored Stop: The robot stops all motion when a human enters the collaborative workspace but remains powered. The stop condition is safety-rated (PL d/e, Cat. 3/4). The human can then perform tasks within the static robot's envelope before restarting it.

Hand Guiding: The human directly physically interacts with the robot to guide its motion, typically via a force-torque sensor or compliant control. The robot only moves when the operator applies force to a dedicated enabling device, providing proportional control for tasks like kinesthetic teaching.

The core of pHRI safety is preventing pain or injury from mechanical contact. ISO/TS 15066 provides biomechanical limit curves based on extensive human subject testing.

• Transient Contact: Brief, impact-like contact. Limits are defined by maximum force (e.g., 190 N for the forehead) and maximum pressure.
• Quasi-Static Contact: Prolonged, clamping, or trapping contact. Limits are significantly lower (e.g., 65 N for the forehead) as pain sensation increases with duration.
• Body Region Specificity: The hand can tolerate higher forces (140 N transient) than the abdomen (130 N transient). Engineers must design for the worst-case body part that could make contact.

Achieving these safety modes requires a integrated stack of hardware and software.

• Safety-Certified Hardware: PL e (Performance Level e) or SIL 3 (Safety Integrity Level 3) rated controllers, drives, and sensors are mandatory for safety functions.
• Dual-Channel Safety Controllers: Use redundant processing to detect faults and initiate safe states.
• Specialized Sensors: Safety-rated vision systems (e.g., PILZ SafetyEye), capacitive skin sensors on the robot, and force/torque sensors at the wrist or joints.
• Real-Time Control Loops: High-frequency impedance control algorithms adjust actuator torque within microseconds to maintain compliant, force-limited interaction.

Impedance Control is a fundamental robotics control strategy for pHRI that regulates the dynamic relationship between a robot's motion and the contact forces it exerts. Instead of dictating a precise position, it controls the robot's apparent mechanical impedance—its stiffness, damping, and inertia—to achieve compliant, natural-feeling interaction.

• Mechanism: The controller modulates actuator torque based on the error between desired and actual position/force, allowing the robot to 'give way' upon contact.
• pHRI Role: Essential for physical guidance (kinesthetic teaching) and safe operation in Power and Force Limiting (PFL) modes. It enables robots to absorb unexpected impacts and collaborate on tasks requiring force exchange, like co-carrying an object.

Admittance Control is a complementary force-control architecture to impedance control, commonly used in pHRI for robots with high inherent stiffness (like many industrial arms). It allows a stiff robot to behave in a compliant manner by using force sensor feedback to modify its commanded motion.

• How it Works: An outer control loop uses measurements from a force/torque sensor at the robot's wrist to calculate a motion correction. This desired velocity or position adjustment is then sent to the robot's fast, inner position controller.
• Comparison to Impedance Control: While impedance control directly modulates torque for compliance, admittance control achieves compliance by mapping sensed forces to motion commands. It is often preferred for precise robots that already have high-gain position controllers.

Kinesthetic Teaching, also known as Lead-Through Programming, is an intuitive robot programming method fundamental to pHRI. A human operator physically grasps the robot's end-effector or arm and manually guides it through the desired task motions, which the robot records and can later replay autonomously.

• Process: The robot is typically placed in a zero-gravity or compliant mode (enabled by impedance/admittance control) to allow effortless guiding. Trajectory waypoints and sometimes forces are captured during demonstration.
• Advantages: Dramatically faster and more accessible than traditional code-based programming, enabling rapid task setup and reprogramming by shop-floor workers.
• Foundation: Relies entirely on safe, responsive physical interaction, making it a quintessential pHRI application.

Shared Autonomy is a control paradigm for pHRI where control authority over a robot's actions is dynamically and intelligently shared between a human operator and an autonomous controller. The system blends human judgment and intuition with machine precision and consistency.

• Dynamic Allocation: The level of autonomy can change based on context, task phase, or user input. For example, the robot may handle fine, tremor-free positioning while the human provides high-level directional guidance.
• Implementation: Often uses techniques like virtual fixtures (software-defined guidance constraints), intent recognition, and adaptive impedance control to smoothly merge human and robot inputs.
• Use Case: Critical in complex manipulation tasks (e.g., surgery, assembly) and assistive robotics, where full autonomy is unreliable but pure teleoperation is inefficient.