VisualProHumanoid Robot FMEA Technical Report
Integrated Analysis of Design Safety & Functional Safety based on VisualPro
Analysis Tool: VisualPro DFMEA | Methodology: AIAG-VDA 7-Step | Date: 2026-07-03
ISO 13482ISO 13849-1 eIEC 61800-5-2 STOISO 26262 ASIL
01Why Robot Safety Now
Humanoid robots are rapidly expanding into manufacturing, logistics, and service environments. As they cooperate with humans in the same workspace, “safety” determines product competitiveness just as much as “precision of movement.”
FMEA (Failure Mode and Effects Analysis), which predicts and controls failures in the early design phase, is the most reliable way to reduce recall and personal injury risks and accelerate functional safety standard certification. Based on actual data from a humanoid robot DFMEA performed with VisualPro, this report presents the risks of four core structures and the direction of safety design.
1 hour in the design phase saves 100 hours in the field.
02VisualPro Performed FMEA — At a Glance
On VisualPro's System Breakdown Tree, the robot was decomposed into four core structures (Brain, Cerebellum, Hand, and Joints), and the Function → Failure Mode (FM) → Failure Cause (FC) → Higher-level Effect (FE) of each structure were connected into a network. Below are the actual analysis tree and the Action Priority (AP) for each structure.
VisualPro FMEA System Breakdown Tree — Action Priority (AP) by structure
03Structure Analysis
The robot was decomposed into the three AIAG-VDA levels (System → Subsystem → Component).
| Level | Structural Element | Core Role |
|---|
| System (L1) | Humanoid Robot | Safely interact with humans and execute missions |
| Subsystem (L2) | ① Brain (Perception & Decision) | Vision, voice, environmental perception, AI decision-making |
| ② Cerebellum (Motion Control) | Posture & balance estimation, real-time gait control |
| ③ Dexterous Hand (Precision Manipulation) | Multi-DOF grasping, force/torque control |
| ④ Joints (Joint Actuation) | Joint torque/position actuation, force/torque feedback |
| Component (L3) | 9 Components | Vision/sensor fusion, AI inference SW, balance estimator, motion controller, F/T sensor, finger actuator, DC servo motor, encoder, motor driver |
04Function & Failure Analysis
Defined the failure modes that occur when the function of each structure is lost, the final Failure Effect (FE) on humans, and the Severity (S).
| Structure | Failure Mode (FM) | Failure Effect on Human (FE) | Severity (S) |
|---|
| Brain | Failure in obstacle perception/decision | Human-robot collision injury | 9 |
| Cerebellum | Failure in posture/balance control | Robot overturning (fall) injury | 10 |
| Hand | Grasping force control failure (excessive/insufficient) | Dropped objects / hand overpressure injury | 8 |
| Joints | Loss of joint actuation / torque runaway | Robot overturning (fall) injury | 10 |
05Risk Profile (Risk Analysis)
Evaluated the Severity (S), Occurrence (O), and Detection (D) for nine Failure Causes (FC), and automatically calculated the AIAG-VDA Action Priority (AP).
Risk Profile by Failure Cause (S · O · D)
VisualPro DFMEA measured data · 9 Failure Causes
Severity SOccurrence ODetection D
Sensor fusion false positive/negative
Brain High
Unidentified unlearned obstacle
Brain High
IMU state estimation divergence
Cerebellum Low
Control cycle non-compliance
Cerebellum Low
Actuator over-grasping
Hand Low
Winding short / torque loss
Joints Low
Encoder signal loss
Joints Low
Driver overcurrent
Joints Low
| Structure | Failure Cause (FC) | S | O | D | AP | Safety Function (Optimization) |
|---|
| Brain | Sensor fusion false positive/negative | 9 | 4 | 5 | High | Speed & Separation Monitoring (SSM), sensor triplication |
| Brain | Unidentified unlearned obstacle | 9 | 4 | 5 | High | Safe-stop fallback, scenario expansion |
| Cerebellum | IMU state estimation divergence | 10 | 3 | 4 | Low | IMU triplication, protective stop |
| Cerebellum | Control cycle non-compliance | 10 | 3 | 3 | Low | Watchdog → STO integration (PLr e) |
| Hand | F/T sensor drift | 8 | 3 | 4 | Low | PFL force limiting, dual F/T sensors |
| Hand | Actuator over-grasping | 8 | 3 | 4 | Low | Torque limiter + PFL |
| Joints | Winding short / torque loss | 10 | 2 | 4 | Low | STO, insulation diagnosis |
| Joints | Encoder signal loss | 10 | 3 | 4 | Low | Dual encoder mismatch → STO |
| Joints | Driver overcurrent | 10 | 3 | 3 | Low | Hardware STO, overcurrent trip |
Interpretation Failure items that lead to overturning or collision have high severity (S) scores of 8–10, but they are managed as mostly AP Low because preventative and detection designs have been implemented. In contrast, two cases in the perception system (Brain) are evaluated as AP High due to the high difficulty of detection (D5) — marking them as the highest priority for optimization.
06Functional Safety Standard Compliance
Humanoids that physically interact with humans apply a layered standard framework rather than a single standard.
| Classification | Standard | Role |
|---|
| Product Safety | ISO 13482 | Safety requirements for personal care robots |
| ISO/TS 15066 | Human-robot contact force/pressure limits (PFL basis) |
| Functional Safety | ISO 13849-1 (PLr d~e) | Performance level of safety-related control systems |
| IEC 61800-5-2 (STO) | Safe Torque Off function |
| Drive ECU | ISO 26262 Adaptation (ASIL C~D) | Automotive-grade actuator electronics safety |
07Key Optimization Measures
Joint ActuationISO 26262 ASIL D
Torque loss or runaway is physically blocked by Safe Torque Off (STO), dual encoders, and overcurrent trip
Motion ControlISO 13849-1 PLr e
RT-OS watchdog timer is integrated with STO, and posture estimation divergence monitoring prevents tipping over
Precision ManipulationISO/TS 15066
Human contact overpressure is prevented by Power and Force Limiting (PFL) and dual F/T sensor cross-validation
Perception & DecisionHighest Priority
Collision avoidance is enhanced through Speed and Separation Monitoring (SSM), safe-stop fallback, and improved detection of unlearned situations
08Summary & Proposal
VisualPro decomposes the robot into structural units and quantifies risk by connecting functions, failures, causes, and effects into a single network. As a final result, it clearly highlights “where, why, and with which safety function” to reinforce the system.
Deconstruct by structure, prove by standard — Robot safety design with VisualPro.
VisualPro & FMEA Consulting Inquiries
Humanoid robots are rapidly expanding into manufacturing, logistics, and service environments. As they cooperate with humans in the same workspace, “safety” determines product competitiveness just as much as “precision of movement.”
FMEA (Failure Mode and Effects Analysis), which predicts and controls failures in the early design phase, is the most reliable way to reduce recall and personal injury risks and accelerate functional safety standard certification. Based on actual data from a humanoid robot DFMEA performed with VisualPro, this report presents the risks of four core structures and the direction of safety design.
On VisualPro's System Breakdown Tree, the robot was decomposed into four core structures (Brain, Cerebellum, Hand, and Joints), and the Function → Failure Mode (FM) → Failure Cause (FC) → Higher-level Effect (FE) of each structure were connected into a network. Below are the actual analysis tree and the Action Priority (AP) for each structure.
The robot was decomposed into the three AIAG-VDA levels (System → Subsystem → Component).
Defined the failure modes that occur when the function of each structure is lost, the final Failure Effect (FE) on humans, and the Severity (S).
Evaluated the Severity (S), Occurrence (O), and Detection (D) for nine Failure Causes (FC), and automatically calculated the AIAG-VDA Action Priority (AP).
Humanoids that physically interact with humans apply a layered standard framework rather than a single standard.
VisualPro decomposes the robot into structural units and quantifies risk by connecting functions, failures, causes, and effects into a single network. As a final result, it clearly highlights “where, why, and with which safety function” to reinforce the system.