Rehabilitation Robot Gait Training Adaptive Control 2026
Rehabilitation Robot Gait Training Adaptive Control
Adaptive control for rehabilitation robots integrates impedance modulation, EMG-driven intent recognition, and reinforcement learning to personalize gait therapy in real time. This dataset spans 50+ patent and literature records from 2007 to 2026.
Adaptive Control Architectures in Rehabilitation Robotics
Rehabilitation robot gait training adaptive control integrates mechanical design, multi-modal sensing, computational intelligence, and human-robot interaction theory into systems that continuously adjust training parameters to match patient capability. Core sub-domains include impedance and admittance control with variable stiffness, EMG-driven intent recognition, and reinforcement learning-based parameter optimization.
The field rests on a recognition that rigid, pre-programmed trajectories yield suboptimal neuroplastic outcomes. Key technical challenges concentrate on three axes: decoding human movement intention in real time, evaluating rehabilitation progress objectively, and selecting control strategies that balance assistance with patient-driven effort, as articulated in the 2022 review on active intelligent gait training systems.
Lower limb and gait-focused systems dominate the dataset, appearing in roughly two-thirds of retrieved records, with upper limb systems comprising the remaining third. Both domains share control architecture patterns — impedance controllers, sEMG-guided gain tuning, assist-as-needed logic — but differ in mechanical configuration and clinical targets including stroke, spinal cord injury, cerebral palsy, and Parkinson’s disease.
Among formal patent records retrieved in this dataset, 6 of 10 filings are dated 2023–2026, confirming accelerating IP activity. Innovation in this dataset is distributed across many institutions rather than concentrated in a few large corporations, with Chinese academic institutions and hospitals among the most active recent filers in retrieved records.
Patent Filing Trends and Jurisdiction Breakdown
Among formal patent records retrieved, CN accounts for 6 filings and US accounts for 5 filings, with 1 EP filing. Filing density in this dataset increases sharply after 2020, with 6 of 10 formal patent records dated 2023–2026.
Patent Filings by Jurisdiction (Retrieved Records)
In this dataset, CN filings account for 6 records and US filings for 5 records, with 1 EP record — no JP or KR filings appeared in retrieved records despite strong literature contributions from those regions.
↗ Click bars to explorePatent Filing Activity by Period (Dataset Snapshot)
In this dataset, filing activity accelerates markedly in the 2023–2026 period, with 6 of 10 formal patent records concentrated in those years, compared to earlier periods.
↗ Click bars to exploreKey Clinical Application Areas in Gait Rehabilitation Robotics
The dataset covers five primary clinical application domains ranging from stroke and spinal cord injury rehabilitation to pediatric neurological disorders, elderly care, and prosthetics. Each domain presents distinct control requirements and patient populations.
Stroke Lower and Upper Limb
The largest application sector in this dataset, addressed by overground exoskeletons, treadmill-based systems, and end-effector platforms. A 2017 survey covers drive modes, training paradigms, and gait detection techniques across platforms. A 2013 study with end-effector robot training demonstrates statistically significant gait speed and stride length improvements in Parkinson’s disease patients. Adaptive admittance control with VR environments was applied to stroke survivors in a 2021 study.
Neurological RehabilitationSpinal Cord Injury Gait Training
Multiple records specifically address incomplete SCI. A 2018 study applies volition-adaptive control modifying joint impedance in real time based on neural signals and interaction torques in incomplete SCI subjects. A 2014 explorative trial with 10 subjects reports improved gait ability using impedance-controlled robotic gait training in individuals with chronic incomplete spinal cord injury.
Spinal Cord InjuryPediatric Cerebral Palsy Therapy
A 2022 study reports significant improvements in GMFM scores and gait speed across 17 pediatric patients using a joint-torque-assisting wearable exoskeletal robot for overground gait training in children with static brain injury. A 2018 study proposes a structured 16-session protocol using the CPWalker platform aligned to ICF-CY framework goals for cerebral palsy rehabilitation.
Pediatric NeurologicalElderly Care Mobility Assistance
Two Chinese patents from Jiangsu Institute of Commerce and Technology address elderly-specific adaptive training control via server-connected rehabilitation robots with weighted feedback matching against reference databases, filed in 2020 and 2022. A 2026 CN patent from Southwest Medical University Affiliated Hospital integrates variable-universe fuzzy reasoning and dynamic risk mapping for patient-specific adaptive assistance compensation.
Elderly RehabilitationKey Patent Assignees in Rehabilitation Robot Adaptive Control (Retrieved Records)
In this dataset, patent filings are distributed across seven named assignees rather than concentrated in a single institution. Chinese academic and hospital-affiliated organizations account for the majority of CN filings in retrieved records, while CUREXO (KR) is pursuing US market protection with a 2026 pending application.
Top Assignees by Filing Count in Retrieved Records (Dataset Snapshot)
↗ Click bars to exploreSoutheast University
Southeast University holds 2 active US patents filed in 2024, both covering adaptive control methods and systems for upper limb rehabilitation robots based on game theory and surface electromyography. The patents encode a Back Propagation Neural Network (BPNN) muscle force model combined with game-theoretic human-robot interaction analysis to modulate control commands. This cross-border filing strategy signals deliberate US jurisdiction IP positioning by a Chinese academic institution.
China — CN (US filings)Jiangsu Institute of Commerce and Technology
Jiangsu Institute of Commerce and Technology (江苏经贸职业技术学院) holds 2 CN patents on server-networked elderly rehabilitation training control, filed in 2020 and 2022, both covering training control methods and devices for elderly rehabilitation robots using weighted feedback matched against reference databases. Both patents are now listed as inactive in the retrieved records.
China — CNConverging Technology Trends in Adaptive Gait Rehabilitation (2024–2026)
The most recent filings in this dataset (2024–2026) point to five converging directions: dynamic risk mapping with neuro-musculoskeletal signal fusion, force-based real-time gait phase estimation, robust constraint trajectory tracking, EtherCAT-based distributed control architecture, and sEMG combined with game-theoretic co-optimization.
Dynamic Risk Mapping and Neuro-Musculoskeletal Signal Fusion
The 2026 CN patent from Southwest Medical University Affiliated Hospital integrates multimodal physiological data streams including motion data and EMG, neuro-muscular cooperative motion state analysis, dynamic obstacle-aware risk mapping, and variable-universe fuzzy reasoning into a unified adaptive assistance compensation system. This represents a convergence of previously separate research threads addressing intent recognition, safety monitoring, and real-time control adaptation in a single system architecture.
Force-Based Real-Time Gait Phase Estimation Without Motion Capture
The 2026 pending US patent from CUREXO introduces footplate-mounted sensors that compute anteroposterior forces per foot under load versus no-load conditions to determine gait status in real time, enabling trajectory and speed adaptation without requiring external motion capture equipment. This approach reduces clinical setup complexity while maintaining continuous gait phase feedback for the adaptive controller.
Impedance/Admittance Control vs. Assist-As-Needed Control: Key Dimensions
Click any row to explore further.
| Dimension | Impedance / Admittance Control | Assist-As-Needed (AAN) Control |
|---|---|---|
| Dataset Record Count | 15+ retrieved records | 12+ retrieved records |
| Core Mechanism | Modulates robot stiffness and damping in response to interaction force; creates compliant virtual channel around reference trajectory | Minimizes robotic assistance to exactly what is needed for task completion; maximizes patient active neural engagement |
| Adaptive Extension | Variable stiffness extensions tighten or loosen virtual channel in real time based on patient performance metrics; RL used to reshape impedance landscape phase-dependently | Performance monitoring of trajectory deviation, velocity, and force dynamically adjusts assistance gain or virtual channel stiffness |
| Signal Integration | Interaction force measurements; sEMG-derived variable impedance (outer loop); sliding mode iterative learning control (inner loop) | Positional error for corrective assistance; fault-tolerant region with stiffness-field gradients; Gaussian Mixture Models for 3D trajectory encoding |
| Clinical Targets | Stroke, spinal cord injury, incomplete SCI, transfemoral amputees, pediatric static brain injury | Stroke upper limb, upper extremity rehabilitation across passive, assistant, active, and resistive modes |
| Representative 2022+ Work | PI²-based impedance landscape shaping (2022); sEMG gain-tuned compliance control (2022); adaptive admittance with VR for stroke (2021) | AAN with Gaussian Mixture Models for bilateral upper limb robot (2022); spatial freedom controller with virtual channel (2020) |
| Operational Mode Range | Passive to active; variable channel width; can incorporate resistive loading | Passive, assistant, active, and resistive modes explicitly integrated in single controller framework |
| RL Integration | Directly integrated — PICE for prosthetic knee, PI² for gait impedance landscape shaping | Less direct RL integration in retrieved records; primarily relies on performance-monitoring feedback loops |
Frequently Asked Questions: Rehabilitation Robot Gait Training Adaptive Control
Impedance control is the dominant paradigm in this dataset, appearing across at least 15 retrieved records. The core mechanism modulates robot stiffness and damping in response to interaction force measurements, creating a compliant virtual channel around a reference trajectory. Variable stiffness extensions allow the controller to tighten or loosen the channel in real time based on patient performance metrics.
sEMG is cited in at least 10 records in this dataset as the dominant biological signal for intent recognition and control adaptation. It maps muscle activation patterns to motion intent or force estimates, feeding these into impedance, compliance, or trajectory correction loops. Applications include normalized sEMG-to-gain mapping for compliance modulation, BPNN-based muscle force estimation combined with game-theoretic interaction analysis, and sEMG-derived variable impedance as the outer loop in dual closed-loop control strategies.
AAN is the most clinically oriented control philosophy in the dataset, appearing in over 12 records. The mechanism minimizes robotic assistance to exactly what is needed for task completion, thereby maximizing the patient’s active neural engagement. Operationally it is implemented via performance monitoring of trajectory deviation, velocity, and force that dynamically adjusts assistance gain or virtual channel stiffness. Implementations span spatial freedom controllers, fault-tolerant stiffness-field gradients from passive to resistive modes, and Gaussian Mixture Models for 3D trajectory encoding.
In this dataset, Southeast University (CN) and Jiangsu Institute of Commerce and Technology (CN) each have 2 patent records. Southeast University holds 2 active US patents from 2024 on sEMG and game-theoretic adaptive control for upper limb robots. Jiangsu Institute of Commerce and Technology holds 2 CN patents (2020 and 2022) on server-networked elderly rehabilitation training control, both now inactive. The remaining five named assignees each have 1 record in retrieved records.
A growing cluster of 8 retrieved records deploys RL as the optimization engine for control parameter adaptation. Approaches include Policy Improvement with Path Integrals (PI²) for phase-dependent impedance landscape shaping, Policy Iteration with Constraint Embedded (PICE) for data-efficient prosthetic knee impedance tuning, and actor-critic heuristic dynamic programming for tracking intact knee profiles in transfemoral amputees. The common formulation treats rehabilitation assistance scheduling as a reward-maximization problem that learns from interaction data without manual calibration.
The most recent filings in this dataset (2024–2026) point to five converging directions: dynamic risk mapping with neuro-musculoskeletal signal fusion (Southwest Medical University Affiliated Hospital, CN 2026); force-based real-time gait phase estimation using footplate-mounted sensors without motion capture (CUREXO, US 2026); robust constraint trajectory tracking matched to patient rehabilitation state (Hefei Deep Valley Technology, CN 2025); EtherCAT bus-based distributed control architecture with six-dimensional force sensing (Beijing Hangrui Kang Technology, CN 2025); and sEMG combined with game-theoretic co-optimization filed in the US by a Chinese academic institution (Southeast University, US 2024).
Data and insights on this page are based on a limited patent and literature dataset and are for reference only. Figures may not represent the complete technology landscape.