TRIAGE: Type-Routed Interventions via Aleatoric-Epistemic Gated Estimation in Robotic Manipulation and Adaptive Perception -- Don't Treat All Uncertainty the Same

Abstract

Most uncertainty-aware robotic systems collapse prediction uncertainty into a single scalar score and use it to trigger uniform corrective responses. This aggregation obscures whether uncertainty arises from corrupted observations or from mismatch between the learned model and the true system dynamics. As a result, corrective actions may be applied to the wrong component of the closed loop, degrading performance relative to leaving the policy unchanged. We introduce a lightweight post hoc framework that decomposes uncertainty into aleatoric and epistemic components and uses these signals to regulate system responses at inference time. Aleatoric uncertainty is estimated from deviations in the observation distribution using a Mahalanobis density model, while epistemic uncertainty is detected using a noise robust forward dynamics ensemble that isolates model mismatch from measurement corruption. The two signals remain empirically near orthogonal during closed loop execution and enable type specific responses. High aleatoric uncertainty triggers observation recovery, while high epistemic uncertainty moderates control actions. The same signals also regulate adaptive perception by guiding model capacity selection during tracking inference. Experiments demonstrate consistent improvements across both control and perception tasks. In robotic manipulation, the decomposed controller improves task success from 59.4% to 80.4% under compound perturbations and outperforms a combined uncertainty baseline by up to 21.0%. In adaptive tracking inference on MOT17, uncertainty-guided model selection reduces average compute by 58.2% relative to a fixed high capacity detector while preserving detection quality within 0.4%. Code and demo videos are available at https://divake.github.io/uncertainty-decomposition/.

Figures (4)

• Figure 2: Empirical behavior of uncertainty signals on MOT17. (a) Aleatoric uncertainty correlates with prediction error, indicating sensitivity to observation corruption. (b) Epistemic uncertainty exhibits a broader spread reflecting variation in representation support. (c) Joint distribution of aleatoric and epistemic signals across 21,324 detections shows near-zero correlation ($r=0.048$), confirming that the two uncertainties capture distinct disturbance mechanisms.
• Figure 3: (a) Aleatoric and epistemic trigger rates under four perturbation conditions ($\pm1$ std over 1,000 episodes). $\sigma_{\text{alea}}$ fires on 85.2% of steps under sensor perturbation while $\sigma_{\text{epis}}$ fires on 12.4% under dynamics shift; under compound perturbation both rates persist (84.6%, 13.8%), confirming near-independence. (b) Object height in a representative compound episode: Vanilla and Total-U fail to lift the cube above the 0.2 m success threshold, while Decomposed succeeds.
• Figure 4: Adaptive model selection for MOT17-04. Top: Temporal uncertainty evolution (epistemic in red, aleatoric in blue). Bottom: Selected models color-coded by capacity (green=Nano, yellow=Small, orange=Medium, red=Large, dark red=XLarge). Policy correlates model scaling with epistemic spikes while ignoring aleatoric elevation, validating orthogonality-aware learning.
• Figure 5: Ablation study comparing our orthogonal uncertainty decomposition against total uncertainty baseline across five MOT17 sequences. Our method achieves 58.2% average computational savings vs. 44.6% for total uncertainty (+13.6% improvement). This substantial gap validates the necessity of separating aleatoric and epistemic components: total uncertainty conservatively uses larger models when combined uncertainty is high, while our method recognizes that high aleatoric uncertainty alone does not require increased model capacity.