Robotics Intelligence Explorer
Explore our educational framework for industrial robotics intelligence: from passive mechanical transmission (Level 0) and high-speed current loops (Level 1), through sensing (Level 2), deterministic kinematic control (Level 3), machine vision perception (Level 4), and adaptive planning (Level 5).
INDUS 6-Tier Robotics Intelligence Framework
This 6-tier taxonomy (Level 0 through Level 5) is an educational engineering framework established by INDUS Industrial Robotics to clarify how physical transmission, electromechanical actuation, deterministic real-time control, and modern AI perception interact. It is designed to assist system integrators, robotics engineers, and equipment purchasers in scoping system complexity.
This framework is not an ISO/IEC standardized classification. Certified safety circuits (ISO 13849-1 / IEC 62061) operate as independent, hardware-level supervisors across all six tiers.
Interactive Robotics Intelligence Stack
Click on any tier below to inspect its mechanical physics, signal interfaces, latency budgets, and real-world failure dependencies.
Hardwired dual-channel safety relays, light curtains, emergency stops, and Safe Torque Off (STO) override software commands at every tier.
Mechanical Transmission & Structural Rigidity
Pure physics, mechanical force transmission, and zero algorithmic computation
Transmits mechanical power, scales rotational speed and torque, supports radial/axial shaft moments, and determines the fundamental geometric precision and mechanical backlash of the system without software intervention.
Deterministic mechanical physics. Governed by Hertzian contact stresses, elastic tooth deflection, friction coefficients, and thermal expansion. Has zero latency and infinite mechanical reliability up to yield stress.
- Rigid position holding at wrist joint against 250 kg gravity moment
- Zero-backlash reversal during micro-dispensing adhesive bead paths
- Absorbing 400% momentary shock impact on spot welding gun closure
| Signal Name | Signal Type | Bandwidth / Rate | Protocol / Interface |
|---|---|---|---|
| INMechanical Input Torque & Angular Velocity | Mechanical Power (N·m, rad/s) | Instantaneous physical transmission | Physical shaft key / clamping hub coupling |
| OUTReduced Speed, Multiplied Output Torque & Reaction Moments | Mechanical Kinetic Output (N·m, N) | Instantaneous physical transmission | Output flange / solid shaft DIN ISO 9409-1 |
- Input State:
- None (passive mechanical structure)
- Processing Core:
- None (strain and gear mesh physics)
- Physical Actuator Output:
- Transmitted torque to robot link arm
- Latency Budget:
- 0 µs (speed of sound in steel ~5,000 m/s)
- Strain-wave flexsplines & circular splines
- RV cycloidal pin-wheel assemblies
- Ground helical and planetary gear stages
- Precision preloaded crossed-roller bearings
- Elastomer and disc pack backlash-free couplings
- Linear guideways and ground ball screws
- • Cannot compensate for mechanical manufacturing tolerances or wear through software
- • Subject to torsional wind-up under dynamic torque acceleration spikes
- • Thermal expansion alters tooth backlash and grease viscosity over extended shifts
Provides mechanical holding torque during e-stop; fail-safe electromagnetic holding brakes physically lock the shaft independent of software state.
Distinguishing Control, Perception & Autonomy
Marketing narratives often blur the boundaries between mathematical feedback control, deep learning vision, and real-time motion generation. Compare their governing principles and safety roles:
| Control Concept | Governing Principle | Typical Frequency | Determinism Guarantee | Replaces Safety Circuit? | Industrial Automation Role |
|---|---|---|---|---|---|
| Deterministic Kinematic Control | Pre-computed trajectory math, inverse Jacobian kinematics, linear/spline interpolation. | 1 kHz (1,000 Hz) fixed cycle | Strict mathematical guarantee: maximum jitter < 1 µs under real-time OS. | ✕ Never (Strictly No) | Interpolating robot wrist paths during precision MIG/laser welding and dispensing. |
| Closed-Loop Feedback Control | Proportional-Integral-Derivative (PID), state-space observers, field-oriented current control. | 10 kHz to 32 kHz current loop; 1 kHz position loop | High determinism within servo drive hardware; instantaneous response to tracking error. | ✕ Never (Strictly No) | Holding commanded motor position against external load disturbance and gravity moments. |
| AI Perception & Segmentation | Deep neural networks (CNNs, Vision Transformers, PointNet) processing raw sensor frames. | 10 Hz to 60 Hz inference rate | Probabilistic: outputs confidence scores, processing time varies with scene complexity. | ✕ Never (Strictly No) | Identifying picking points for random cast parts in deep logistics totes. |
| Adaptive Impedance Planning | Online optimization, Model Predictive Control (MPC), dynamic admittance adaptation. | 50 Hz to 500 Hz re-planning rate | Bounded optimization time with fallback default behaviors if solver does not converge. | ✕ Never (Strictly No) | Regulating contact pressure during composite polishing without gouging workpiece. |
| Autonomous Mission Decision-Making | Behavior trees, high-level task planners, fleet dispatching, semantic SLAM. | 0.1 Hz to 5 Hz asynchronous decisions | Heuristic and probabilistic; relies on lower deterministic layers for physical execution. | ✕ Never (Strictly No) | Routing 40 AMRs across a 50,000 m² warehouse to optimize order picking throughput. |