Why Position Control Is Fundamentally Unsafe for pHRI Contact Events

Position control commands a robot manipulator to track a specified trajectory by driving joint angle errors to zero — producing high stiffness to resist any deviation, including an unexpected human body. This structural property, useful in isolated industrial cells, becomes a direct hazard in shared human-robot workspaces: the robot treats human contact as a disturbance to be overcome, potentially exerting unconstrained forces. As documented in Sony’s 2012 patent on robot control methods, traditional position-controlled robots “provide angle command values to joint actuators and drive joints strictly to those values,” making them “weak in soft control of force or acceleration.”

FANUC’s 2012 collaborative robot patent reinforces this structural weakness. Its architecture delineates a “first robot section” far from the human — potentially dangerous if position-controlled at speed — and a “second robot section” near the human that is force-limited. The design implicitly acknowledges that position-controlled segments remain hazardous unless physically separated from the human workspace. The safety strategy is spatial segregation, not contact management.

Panasonic’s 2008 patent on robot control methods takes the avoidance-focused approach characteristic of position-control-era safety: predicting whether the robot will contact a human and restricting motion to avoid contact with vital body spots. This predict-and-avoid paradigm contrasts sharply with impedance control, which manages contact forces during or after contact rather than attempting to prevent contact entirely. In dynamic collaborative environments — assembly lines, surgical assistance, rehabilitation robotics — avoidance-only strategies are operationally insufficient.

Universal Robots’ 2021 safety controller patent illustrates a further limitation of position-control-era safety: its “violation stop mode” halts the arm when operating parameters exceed defined limits. While adaptive in its threshold-setting, this remains a hard-limit enforcement model — it does not modulate the robot’s compliance during contact. Critically, abrupt stops can themselves be hazardous if a human is physically entangled with the arm when the safety trigger fires, a scenario that impedance control can handle gracefully by yielding rather than halting.

The patent landscape, as analysed by PatSnap’s innovation intelligence platform, confirms that position control safety architectures rely on discrete binary decisions — stop or continue — rather than the continuous force modulation that pHRI safety demands. Standards bodies including ISO (specifically ISO/TS 15066 on collaborative robots) and IEC have increasingly codified force and pressure limits for human-robot collaboration, reflecting the industry’s recognition that position-only control is insufficient for shared workspaces.

Impedance Control: Governing Forces Through Virtual Mechanical Dynamics

Impedance control reframes the pHRI safety problem entirely by regulating the dynamic relationship between a robot’s position deviation and the forces it exerts, modeled as a virtual mass-spring-damper system. When a human contacts an impedance-controlled robot, the robot yields compliantly, dissipating kinetic energy through virtual damping rather than resisting the interaction. Force management is embedded in the continuous control law itself — not added as an external watchdog check — which is the architectural distinction that makes impedance control categorically different from position control for pHRI safety.

GM Global Technology Operations’ 2013 patent on workspace-safe force/impedance-controlled robots establishes the foundational architecture: a saturation limit on the static force applied by the manipulator, combined with a dynamic reflex that calculates the required reflex torque at the joint actuator when contact force exceeds a threshold. This dynamic reflex is critical for pHRI — it addresses the inertial impulse that static force saturation alone cannot handle, since impact transients can cause injury within milliseconds. No equivalent mechanism exists within a pure position-control framework.

“In cluttered or unknown environments where collisions are unavoidable, joint-space safety should be achieved not by collision avoidance but by limiting contact force magnitude — a goal that is naturally expressed and enforced through impedance control but not through position control.”

Franka Emika’s 2023 force-limitation patent specifies a maximum permissible force and uses impedance regulation — computing the current reference force of an artificial spring component based on spring stiffness and the difference between current and target position — to ensure this force is never exceeded. If the reference force from the spring exceeds the permissible maximum, an emergency control program is triggered. The key architectural insight is that force management is embedded in the continuous control law, not bolted on as an external watchdog.

Franka Emika extends this architecture further into anatomically-aware safety in its 2022 and 2023 patents on control of robot manipulators in contact with humans. These provide a database of human body zones, each assigned a maximum allowable contact pressure. Upon detecting or predicting contact with a specific body zone, the controller determines a fixed reference position relative to the human body — indicating the start of tissue indentation — and uses this as the zero position for the impedance spring component. The spring then governs force as a function of tissue deformation rather than robot-frame displacement, directly preventing the robot from exceeding biomechanically safe tissue pressure limits. This level of specificity is documented by ISO/TS 15066 as the correct approach for collaborative robot safety, and it is architecturally impossible in position control.

Panasonic’s 2009 patent on robot arm control introduces a danger-degree calculation based on the relative position of the robot arm and the collaborating human, using this danger degree to set the mechanical impedance value of the arm through an impedance setting unit. Notably, the patent specifies that the higher the danger degree, the higher the stiffness set for the robot arm — counterintuitively increasing rigidity when proximity is close. This approach uses impedance as a spatial safety gating function, blending position-awareness with force-law modulation in a way that pure position control cannot achieve without separate force-limiting hardware.

KUKA’s 2018 multi-robot group control patent makes this force-velocity tradeoff explicit: the system allows arbitrary speeds at low forces and arbitrary forces at low speeds. This parameterization only makes sense within a force/impedance control framework — in position control, force is not a directly commanded or regulated variable, so such a tradeoff cannot be expressed in the control law itself.

Adaptive and Intent-Driven Impedance Control for Dynamic pHRI Safety

Classical position control and fixed-parameter impedance control share a critical limitation: neither can adapt to the changing dynamics of human intent and task state. Recent patents demonstrate sophisticated schemes for dynamically tuning impedance parameters in real time based on estimated human motion intent, collaboration state, and task context — closing the gap between task performance and safety that fixed-parameter approaches cannot bridge.

The Chinese Academy of Sciences Institute of Automation’s 2025 patent on human-intent-estimation-based impedance adjustment fuses multimodal sensor signals from the robot end-effector — position, velocity, acceleration, and contact force — into a human intent estimator. This estimator drives an adaptive force-coupling enhanced Dynamic Movement Primitive module that generates both a desired motion trajectory and stiffness parameters, which are then fed to an auxiliary impedance controller. The system explicitly targets non-structured and irregular surface interaction scenarios, where fixed-parameter approaches would either be too stiff (risking injury) or too compliant (losing task performance).

Tsinghua University’s 2021 patent on intent-driven adaptive impedance control proposes calculating the user’s motion intent through a motion intent model and tracking that intent in real time to acquire variable impedance model parameters. These parameters are fed into the driver’s control system model to adjust actuator output torque. The patent explicitly acknowledges that existing methods — including simple position-following and reactive collision detection — “impose excessive real-time sensor requirements and severely limit performance while ensuring safety,” framing adaptive impedance as the resolution to this tradeoff.

Kwangwoon University’s 2023 patent on variable admittance control based on human-robot collaboration state addresses the dual problem of distinguishing intended from unintended human actions. The approach sensitively sets admittance parameters in safe collaboration states and insensitively sets them in unsafe states, using frequency analysis to classify the collaboration state with high frequency resolution and low computational load. Admittance control — the dual of impedance control, commanding velocity in response to force rather than force in response to position — is particularly relevant where the human directly applies forces to guide the robot, as in collaborative assembly or rehabilitation.

Nanjing Estun Automation’s 2024 patent directly addresses a residual weakness of conventional admittance and impedance control: the inability to satisfy velocity and position safety constraints simultaneously. The proposed method generates desired interaction trajectories from the interaction control equation and interaction forces, then shapes these trajectories according to safety limits to produce a safe reference trajectory delivered to the robot controller. Speed limiting prevents the robot from running too fast during interaction, while position constraints prevent collisions with the environment and ensure the interaction stays within a safe spatial zone. This hybrid approach captures the force-compliance strengths of impedance control while adding the spatial constraint guarantees that position control architectures traditionally offered — representing the current state of the art in pHRI safety architecture, as tracked by PatSnap’s patent analytics tools.

Reinforcement Learning and Variable Impedance for Contact-Safe Operation

Reinforcement learning (RL) integration with variable impedance control represents the leading edge of pHRI safety innovation in the patent record. Classical position control provides no natural interface for RL to express compliant interaction behavior — the learned policy outputs position targets that the stiff controller then tracks blindly. Impedance control, by contrast, provides a parameterized compliance space — stiffness, damping, and inertia — that an RL agent can tune to achieve both task performance and contact safety simultaneously.

Tsinghua University’s 2022 patent on RL methods for guaranteed contact safety uses a reinforcement learning strategy to generate Cartesian-space variable impedance control parameters for the robot’s target task. If a collision occurs, the collision force magnitude is used to compute a variable impedance control compensation amount, adjusting both the impedance parameters and attitude constraints simultaneously to generate a compensated action. The patent argues that in cluttered or unknown environments where collisions are unavoidable, “joint-space safety should be achieved not by collision avoidance but by limiting contact force magnitude” — a goal naturally expressed and enforced through impedance control but not through position control.

South China University of Technology’s 2023 patent on joint-learning-based optimal human-robot interaction impedance control addresses the case where robot dynamics and human motion characteristics are both unknown. It builds a second-order impedance model for the human-robot interaction task space augmented system and uses an integral reinforcement learning algorithm to update impedance model parameters online until optimal parameters are obtained. An adaptive neural network impedance controller implements these parameters, with a constant neural network controller encoding experience-based knowledge for stable closed-loop performance. This optimization-in-the-loop approach to impedance parameter selection is architecturally impossible in position control, which has no equivalent mechanism for expressing interaction compliance as a continuously optimizable quantity. Research published by IEEE on robot learning confirms that variable impedance parameterization is the preferred interface for RL-based contact-safe manipulation.

Beihang University’s 2023 patent extends impedance control into the null space of a 7-DOF manipulator, constructing a spring-damper-stiffness model that governs the position-force relationship during robot motion. The null-space component allows the robot arm to reconfigure around obstacles — including nearby human operators — without altering the end-effector pose. This architecture enables the robot to simultaneously maintain task accuracy at the end-effector through impedance and avoid operator contact through null-space motion, achieving multi-objective pHRI safety in a unified control law. Neither position control nor simple end-effector impedance control can provide this capability.

Canon’s patents on impedance parameter adaptation during direct teaching address a regime where the distinction between impedance and position control is operationally critical: when a human physically moves the robot arm to program trajectories. Canon’s 2025 and 2019 patents reduce impedance parameters when no workpiece contact is detected (making the arm easy to move) and increase them when contact is detected (providing resistance that communicates the contact state to the teacher). This application of variable impedance to human-robot teaching interaction has no equivalent in position control, where the arm’s resistance to being moved is a fixed function of its stiffness and cannot be modulated based on contact state.

DENSO WAVE’s safety-scene-switching architecture, covered in 2024 and 2025 patents, bridges position-aware safety with dynamically applied safety parameter sets. Multiple parameterized safety scenes are tied to robot position conditions — when the robot enters a zone defined by its position, a corresponding set of safety parameters is activated. This architecture combines the spatial awareness of position control (knowing where the robot is) with the dynamic force-management capability of impedance control (what safety constraints apply in each zone), representing a practical convergence of the two paradigms in industrial collaborative robot deployments.

Head-to-Head: Impedance vs. Position Control Across Eight Safety Dimensions

The patent record enables a direct comparison of how impedance control and position control perform across the safety dimensions most critical to physical human-robot interaction. The table below synthesises the key architectural distinctions documented across the 50+ patents analysed.

The fundamental distinction is architectural. Position control treats the robot as a kinematic machine to be kept on a trajectory, with safety achieved by stopping that machine when proximity or force thresholds are violated. Impedance control treats the robot as a compliant mechanical system whose dynamic relationship to the environment is the primary controlled variable. For pHRI, this means impedance control can manage the contact event itself — its magnitude, duration, and spatial direction — rather than simply detecting and reacting to it.

The remaining gap — that impedance control alone cannot guarantee velocity and position constraints — is the active frontier of innovation. Nanjing Estun Automation’s 2024 hybrid trajectory-shaping approach, DENSO WAVE’s safety-scene-switching architecture, and Beihang University’s null-space unified control law all represent different engineering responses to this gap. The patent landscape, as analysed through PatSnap‘s global patent database, indicates that the state of the art is converging toward architectures that combine impedance-based compliance for contact force management with position/velocity constraint enforcement for workspace safety — with WIPO patent filings in this combined category growing consistently from 2019 through 2025.