Introduction
This page presents an uncertainty taxonomy (UncerTax) for self-adaptive robotics, providing a common vocabulary and structured classification to describe where uncertainty arises, how it behaves, and how it impacts robotic systems and their stakeholders. The taxonomy is derived from research on identifying uncertainty in self-adaptive systems, published in IEEE Software [1].
[1] Hassan Sartaj, Jalil Boudjadar, Mirgita Frasheri, Shaukat Ali, and Peter Gorm Larsen, "Identifying Uncertainty in Self-Adaptive Robotics With Large Language Models," IEEE Software, vol. 43, no. 1, pp. 89-97, Jan.-Feb. 2026, DOI: 10.1109/MS.2025.3620578. Link
Taxonomy Dimensions Overview
The following dimensions serve as the interpretive basis for the taxonomy classification presented in the explorer below.
| Dimension | Explanation | Example |
|---|---|---|
| Nature | How uncertainty behaves: Deterministic/Stochastic, Static/Dynamic. | Stochastic: Random sensor noise; Dynamic: Moving target. |
| Type | Nature of uncertainty: Aleatory (randomness) vs Epistemic (lack of knowledge). | Aleatory: Sensor noise; Epistemic: Incomplete map of the environment. |
| Stage | When uncertainty occurs: Design, Development, Testing, or Operation stages. | Design: Incomplete requirements; Operation: Real-time obstacle detection. |
| Temporal | Time characteristics: Short-Term/Transient and Long-Term/Persistent. | Short-Term: GPS signal loss; Long-Term: Sensor drift. |
| Context of Occurrence | Where it arises: Environmental, Task, Interaction, or Mission. | Environmental: Weather; Task: Unclear goals; Interaction: User input. |
| Source of Adaptation | Origin: External (environment) vs Internal (robot). | External: Changing environmental conditions; Internal: Hardware wear. |
| Scope | Where uncertainty impacts: Local/Global, Component/System Level. | Local: Single sensor; Global: Uncertainty affecting the entire robot. |
| Risk/Severity | How risky the uncertainty is: Low/High Risk. | High Risk: Uncertainty in braking during autonomous driving. |
| Affect | What is impacted: Performance, Safety, Adaptability, or Reliability. | Safety: Avoiding collisions; Performance: Task completion. |
| Propagation | How uncertainty spreads: Isolated versus Cascading. | Cascading: Localization error leading to path-planning errors. |
| Resolution | Approach to handling: Reactive/Proactive, Manual/Automated. | Reactive: Robot reacts to obstacles; Proactive: Prediction models. |
| Data Characteristics | Nature of data: Incomplete, Ambiguous, or Noisy. | Noisy: Sensor data errors; Ambiguous: Misinterpreted readings. |
| Ethical Implications | Considerations for trust, transparency, bias, and fairness. | Bias: Incorrectly prioritizing users due to biased training data. |
Uncertainty Taxonomy Explorer
Explore the full dataset. Each item expands to show the complete 13-dimension classification details.