Book a demo

Cut patent&paper research from weeks to hours with PatSnap Eureka AI!

Try now

Cobot Force Torque Control Landscape 2026 — PatSnap Eureka

Cobot Force Torque Control Landscape 2026 — PatSnap Eureka
Tools Explore in Eureka
Reading14 min
PublishedJun 2, 2025
Coverage1990–2026
Technology Landscape 2026

Collaborative Robot Force Torque Control: Patent & Innovation Landscape 2026

Sensor architectures, control algorithms, and safety frameworks are converging to enable cobots to operate safely alongside humans. This report maps 14 patent families and 30+ literature sources spanning ISO/TS 15066-compliant force control from foundational 1990 filings through active 2026 applications.

Fig. 01 — Patent Filings by Jurisdiction
Cobot Force-Torque Patent Filings by Jurisdiction: US 10, CN 8, EP/FR 4, JP 3 Bar chart showing distribution of 14 patent records across jurisdictions in the collaborative robot force-torque control dataset. US leads with 10 filings, followed by CN with 8, EP/FR with 4, and JP with 3. Source: PatSnap Eureka patent dataset. US 10 CN 8 EP/FR 4 JP 3
Published by PatSnap Insights Team··14 min read Verified by PatSnap Eureka Data
Technology Overview

Three Subsystems Define Cobot Force-Torque Control

Collaborative robot force-torque control is defined by the intersection of three technical subsystems: sensing mechanisms that measure joint torques, end-effector forces, or estimate them virtually; control architectures that translate those measurements into safe, compliant motion commands; and safety and response logic that detects collisions, classifies intent, and triggers appropriate robot behavior.

The field is grounded in the principle that robotic joints must either physically measure or computationally estimate forces and torques to enable compliance. Early patents such as the Westinghouse Electric multiaxis robot controller (US, 1992) established the foundational concept of controlling workpoint torque as an end variable rather than an intermediate variable — a distinction that remains central to modern cobot design. Contemporary filings extend this with neural-network-based dynamic modeling, impedance control, and intent-recognition frameworks driven by force/torque data streams.

Among the retrieved results, 14 distinct patent records and over 30 literature sources span these subsystems across jurisdictions including US, EP, JP, CN, FR, and WO. Sub-domains identified include joint torque sensing and virtual sensing, collision detection and mode-switching control, impedance and admittance control, force-based task learning and imitation, multi-robot torque-coordinated cooperation, and safety threshold management under ISO/TS 15066. For broader context on human-robot collaboration standards, the IEC and IEEE publish complementary safety and communication frameworks. PatSnap’s IP analytics platform enables landscape analysis across all these domains.

PatSnap Eureka Dataset covers 14 patent records and 30+ literature sources spanning US, EP, JP, CN, FR, and WO jurisdictions from 1990 to 2026. Explore the dataset ↗
14
Distinct patent records in dataset
30+
Literature sources analyzed
60%
Filings from 2022–2026
6
Patent jurisdictions covered
14
Distinct assignees identified
4
Core technology clusters
Innovation Timeline

From Foundational Torque Control to AI-Integrated Cobots

Patent activity spans three distinct eras — each building on the prior and accelerating toward AI-integrated force sensing architectures.

Foundational Era · 1990–2013

Workpoint Torque as a Controlled Variable

Staubli International AG (EP, 1990) and Westinghouse Electric (US, 1992) established the concept of controlling workpoint torque as an end variable rather than an intermediate variable — a distinction central to modern cobot design. Kabushiki Kaisha Toshiba’s Robot Control Apparatus (US/EP, 2013) introduced external torque estimation through the difference between estimated drive torque and commanded torque, a method still widely applied. Nachi-Fujikoshi’s articulated robot torque control work appeared in Japan in 2007.

Foundational concept still in use
Development Cluster · 2015–2021

Application-Specific Proliferation

General Motors filed force-torque imitation learning patents in CN (2015, 2017). X Development LLC patented programming and execution of force-based tasks with torque-controlled robot arms (US, 2015). CEA filed an H∞-optimized force amplification control patent (FR, 2017). Hanwha Robotics produced a collision detection mode-switching patent family in US and EP (2020). BAE Systems filed a torque-threshold management system across EP (2019) and US (2022). Shanghai Jiao Tong University filed a torque-sensor-based integrated drive-control architecture (CN, 2018).

Application-specific proliferation
Maturation & AI Integration · 2022–2026

Neural Networks, Precision Sensing, Multi-Robot Coordination

Schaeffler Technologies AG filed a method for precisely determining output torque (US, 2026 — the most recent patent in this dataset). Intel Corporation filed a neural-network cobot model generation system (EP, 2025). Nachi-Fujikoshi filed a safety-tiered escape mode system (JP, 2025; US, 2024). Jaka Robot filed a torque inner-loop multi-robot coordination method (CN, 2025). UBTECH Robotics filed a Riemannian Motion Policy-based force control method (US, 2025). Active and pending filings from 2022–2026 constitute approximately 60% of total patent records in this dataset.

~60% of dataset from 2022–2026
Competitive Structure

Highly Fragmented: 14 Distinct Assignees Across 14 Records

Innovation is not concentrated in a few players — 14 distinct assignees appear across 14 patent records, suggesting a highly fragmented competitive field. Chinese academic institutions and robotics startups are among the most active recent filers, while established European industrial companies such as Schaeffler and CEA continue to contribute precision control innovations. This fragmentation signals an open competitive landscape with significant whitespace for new entrants, particularly in multi-robot force coordination.

Highly fragmented field
PatSnap Eureka Timeline derived from patent filing dates across US, EP, CN, JP, FR, and WO jurisdictions. Dataset represents a snapshot, not a comprehensive industry view. Explore timeline ↗
Technology Clusters

Four Control Approach Clusters in the Patent Dataset

Patent and literature evidence clusters around four distinct technical approaches to cobot force-torque control, from hardware sensing to AI-driven intent recognition.

Filing Activity by Era

Active and pending filings from 2022–2026 constitute approximately 60% of total patent records, signalling accelerating innovation.

Cobot Force-Torque Patent Filings by Era: Foundational 1990–2013 ~40%, Development 2015–2021 ~40%, AI Integration 2022–2026 ~60% Donut chart showing approximate share of patent records across three innovation eras. The 2022–2026 AI integration era constitutes approximately 60% of total records. Source: PatSnap Eureka patent dataset. ~60% 2022–2026 AI Integration 2022–2026 Development 2015–2021 Foundational 1990–2013

Cluster Representation in Dataset

Physical joint torque sensing is the dominant approach; AI-driven intent recognition represents the leading edge of active filings.

Cobot Force-Torque Control Clusters: Physical Sensing 5 patents, Virtual/Sensorless 4 patents, Impedance/Admittance 3 patents, Intent Recognition 3 patents Horizontal bar chart showing patent record counts per technology cluster in the collaborative robot force-torque control dataset. Physical joint torque sensing leads with 5 records. Source: PatSnap Eureka. Physical Sensing 5 Virtual / Sensorless 4 Impedance / Admittance 3 Intent Recognition / AI 3
PatSnap Eureka Cluster counts derived from manual classification of 14 patent records. Records may appear in multiple clusters where approaches overlap. Explore the data ↗
Key Technology Approaches

Control Architecture Progression: From Hardware to AI

The four clusters follow a logical progression from physical sensing through virtual estimation to fully AI-driven adaptive control.

Cluster 1 — Physical Sensing
Per-Axis Torque Sensors
Nachi-Fujikoshi (US, 2024): computes “external force torque” as difference between measured and model-estimated torques; implements three-threshold escape mode system.
Threshold Iteration
BAE Systems (US, 2022): iteratively increases torque thresholds to find minimum safe operating margin without exceeding human contact force limits.
Mode-Switching on Collision
Hanwha Robotics (US, 2020): switches between position and torque control modes upon collision detection, applying directional compensation values per joint.
Cluster 2 — Virtual Sensing
High-Precision Sensorless Estimation
Schaeffler Technologies (US, 2026): targets imprecision of motor-current-based estimation while avoiding cost of redundant dual-sensor systems.
ML-Based Torque Observers
Virtual force sensor literature (2023): achieves correlation coefficients of ~0.99 between estimated and actual torques using EtherCAT-enabled hardware.
Unified Drive-Sense-Control
Shanghai Jiao Tong University (CN, 2018): fuses drive, sensing, and control in a single system supporting collision detection and drag teaching from one torque-sensor input.
Unlock Clusters 3 & 4: Impedance Control & AI Intent Recognition
See how CEA’s H∞ admittance optimization, Philips’ intent-recognition system, and UBTECH’s Riemannian Motion Policy approach reshape the force-control landscape.
H∞ AdmittanceIntent RecognitionRMP Force Control+ more
Unlock full analysis →
PatSnap Eureka Control architecture analysis derived from patent claims and literature methods across 14 records and 30+ sources. Explore control approaches ↗
Application Domains

Where Cobot Force-Torque Control Is Being Deployed

DomainKey Patent / SourceAssigneeYearForce Control Role
Industrial AssemblyProgramming and Execution of Force-Based TasksX Development LLC2015Joint torque commands track position and exert prescribed end-effector forces simultaneously
Industrial AssemblyRAM Insertion via Reinforcement Learning (Literature)2021Torque-controlled robot performs contact-rich assembly using proactive visual residual RL
Medical / SurgicalAutomatic Selection of Collaborative Robot Control ParametersKoninklijke Philips N.V.2023Distinguishes intentional user guidance forces from accidental contact in surgical procedures
Medical / SurgicalComputed Torque Control with RBF Neural Networks (Literature)2020Sub-millimeter force-regulated positioning for percutaneous puncture surgery
Unlock full application domain table
See how force-torque control is applied across human-robot co-manipulation, multi-robot cooperation, and semiconductor manufacturing — with specific patent citations.
Co-manipulationMulti-robotSemiconductor+ more
View all domains →
PatSnap Eureka Application domain mapping derived from patent abstracts and claims across 14 patent records in the dataset. Explore applications ↗
Geographic & Assignee Landscape

US Leads in Volume; China Is the Fastest-Growing Origin

Among the 14 patent records with assignee data in this dataset, the United States is the largest single jurisdiction with 10 filings, including active grants and pending applications. Key US assignees include X Development LLC (Google), Hanwha Robotics Corporation, BAE Systems PLC, Koninklijke Philips N.V., Neuromeka, Nachi-Fujikoshi Corp., Schaeffler Technologies AG, Southeast University, UBTECH Robotics Corp., and Dexterity Inc.

China has 8 CN-jurisdiction records, representing the fastest-growing filing base in this dataset. Key CN assignees include Shanghai Jiao Tong University, Ningbo Institute of Materials Technology and Engineering (Chinese Academy of Sciences), Jaka Robot Co., Ltd., KUKA Robot Manufacturing (Shanghai), Tencent Technology (Shenzhen), and Beihang University Hangzhou Innovation Research Institute. IP strategists should treat CN as a primary jurisdiction — freedom-to-operate analysis must be China-first.

Japan has 3 JP-jurisdiction records, with Nachi-Fujikoshi and Yaskawa Electric as primary filers, reflecting Japan’s long-standing position in industrial robot hardware design. Europe includes BAE Systems (EP, 2019), Hanwha Robotics (EP, 2020), Staubli International (EP, 1990), CEA (FR, 2017), and Intel Corporation’s EP filing (2025) — a notable entry of a major semiconductor company. For global patent filing trends, WIPO and EPO publish complementary jurisdiction data. PatSnap’s materials and engineering solutions and customer case studies demonstrate how IP teams use landscape analysis to prioritize jurisdictions.

PatSnap Eureka Geographic analysis based on 14 patent records with assignee data. CN filings represent the fastest-growing base in this dataset. Explore by jurisdiction ↗
10
US patent records (largest jurisdiction)
8
CN records (fastest-growing base)
14
Distinct assignees across 14 records
3
JP records (Nachi-Fujikoshi, Yaskawa)
4
EP/FR records including Intel 2025
6
Total jurisdictions represented
Emerging Directions 2024–2026

Five Signals Visible in the Most Recent Filings

The 2024–2026 filing cohort reveals directional bets being placed by established industrials, academic spinouts, and AI-native companies.

Precision Sensorless Output Torque Determination

Schaeffler Technologies’ 2026 US pending application directly addresses the gap between expensive dual-sensor systems and imprecise motor-current-based estimation, targeting high-performance cobot systems without hardware overhead.

Neural-Network Cobot Model Generation

Intel Corporation’s 2025 EP filing uses a trained neural network to map joint trajectories from a generic robot model to a specific cobot model, enabling generalized force-torque control transfer across cobot morphologies.

Tiered Safety and Escape-Mode Logic

Nachi-Fujikoshi’s 2024 US and 2025 JP filings implement multi-threshold collision response logic that distinguishes between worker entrapment scenarios requiring escape mode activation versus heavy payload scenarios where escape mode would be dangerous — a significant advance in ISO/TS 15066 implementation granularity.

Torque Inner-Loop Multi-Robot Coordination

Jaka Robot’s 2025 CN filing distributes torque-based compliance control across master and slave robots using preset direction matrices, eliminating the need for a high-performance central computing core and lowering the hardware barrier for multi-cobot deployment.

Unlock Direction 5 & Strategic Whitespace Analysis
See UBTECH’s Riemannian Motion Policy force control approach and the multi-robot coordination whitespace opportunity identified in the 2025 filing cohort.
RMP Force ControlMulti-robot whitespaceFiling opportunity
Unlock emerging directions →
PatSnap Eureka Emerging directions derived from filings dated 2024–2026 in the dataset. Signals represent innovation activity within this dataset only. Explore emerging filings ↗
Strategic Implications

Five Decisions IP and R&D Teams Must Make Now

The competitive and technical signals in this dataset have direct implications for product development, IP strategy, and freedom-to-operate analysis.

Cost-Performance Tradeoff

Sensor Hardware vs. Software Estimation Is the Defining Choice

Multiple active filings — Schaeffler 2026, virtual force sensor literature 2023, Shanghai Jiao Tong University 2018 — converge on the challenge of achieving hardware-grade torque accuracy through software means. R&D teams entering this space must choose a position on this spectrum early, as it constrains downstream control architecture choices. PatSnap’s IP analytics tools can map the prior art landscape for each approach.

Architecture-constraining decision
Safety Standards

ISO/TS 15066 Compliance Is Driving Hardware-Software Co-Design

Nachi-Fujikoshi’s escape-mode logic and BAE Systems’ threshold iteration system both embed safety-standard requirements directly into the control law. IP strategists should monitor whether this safety-by-design approach becomes a mandatory qualification barrier in target markets. Relevant standards context is available from ISO and IEC.

Potential qualification barrier
China IP Strategy

China Is an Innovation Origin, Not Just a Manufacturing Destination

With filings from Chinese Academy of Sciences, Shanghai Jiao Tong University, Jaka Robot, Tencent, KUKA Shanghai, and Beihang University all appearing in this dataset, product developers and IP strategists cannot treat CN as a secondary jurisdiction. Freedom-to-operate analysis must be China-first. The WIPO PCT database and PatSnap’s API access support comprehensive CN prior art searches.

China-first FTO required
AI Integration Timeline

AI-Integrated Force Control Expected to Be Table-Stakes by 2027–2028

Koninklijke Philips’ intent-recognition system, GM’s force-torque imitation learning, and Intel’s neural-network cobot model all embed machine learning directly into the force-torque control loop. Teams building next-generation cobots should expect AI-integrated force control to become a table-stakes feature by 2027–2028. Research context is available from IEEE robotics publications.

Table-stakes by 2027–2028
PatSnap Eureka Strategic implications derived from patent claim analysis and filing trajectory across the 14-record dataset. PatSnap’s innovation intelligence platform supports full landscape and FTO analysis. Explore strategic signals ↗
Frequently asked questions

Collaborative Robot Force Torque Control — key questions answered

Still have questions? PatSnap Eureka can answer them instantly from patent and research data. Ask Eureka ↗
PatSnap Eureka

Generate Your Own Cobot Force-Torque Control Landscape

Join 18,000+ innovators using PatSnap Eureka to generate reports like this one for any technology area.

Ask anything about cobot force-torque control.
PatSnap Eureka searches patents and research literature to answer instantly.
Powered by PatSnap Eureka
Link copied to clipboard