Type-Routed Interventions via Aleatoric-Epistemic Gated Estimation

Don't Treat All Uncertainty the Same

1University of Illinois at Chicago, 2Intel Labs
Overview of uncertainty decomposition framework

Our framework decomposes uncertainty into aleatoric (sensor noise) and epistemic (dynamics mismatch) components, enabling type-specific corrective responses for both robotic control and adaptive visual perception.

Robotic Manipulation (Isaac Lab)

Franka Panda lift-cube task under compound perturbations

Franka Panda: grasp-and-lift under simultaneous sensor noise + dynamics shift

Physical Robot + Vision Pipeline

Agilex robot with adaptive model selection

Agilex MiniScout with uncertainty-guided adaptive model selection on edge

+21.0%
Improvement over combined uncertainty under compound perturbations
80.4%
Task success rate under compound perturbations (vs. 44.6% vanilla)
58.2%
Compute savings in adaptive tracking with negligible accuracy loss
r = 0.048
Near-zero correlation confirming orthogonality of uncertainty signals

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%.

Method

Our framework operates as a post-hoc module that requires no retraining of the base policy. It decomposes uncertainty into two physically distinct components:

Aleatoric Uncertainty (Sensor Noise)

Estimated via Mahalanobis distance in observation space. Measures deviation from the nominal state distribution. When triggered, applies observation recovery by resampling the sensor (N=5 resamples, averaging).

Examples: encoder noise, calibration drift, occlusion, blur

Epistemic Uncertainty (Dynamics Mismatch)

Detected using a noise-robust forward dynamics ensemble (K=5 MLPs) trained with noise-augmented data paired with clean targets. When triggered, applies action dampening (scaling factor α=0.30).

Examples: mass change, friction variation, actuator drift, viewpoint shift

A key property is empirical orthogonality: the two signals maintain near-zero correlation (r=0.048 across 21,324 MOT17 detections), confirming they capture distinct disturbance mechanisms. This enables independent, type-specific interventions without cross-contamination.

Results: Robotic Manipulation

Task success rate (%) on Isaac-Lift-Cube-Franka (N=1,000 episodes per condition). The Franka Emika Panda 7-DOF arm must reach, grasp, and lift a cube to a target height.

Method Nominal Sensor Noise Dynamics Shift Compound
Vanilla (no intervention) 100.0 63.8 72.4 44.6
Observation Recovery Only 97.4 89.2 64.8 57.3
Action Dampen Only 88.6 58.4 78.3 50.2
Total-U (combined scalar) 96.8 80.6 70.1 59.4
Decomposed (Ours) 99.6 94.2 84.2 80.4
Δ vs. Total-U +2.8 +13.6 +14.1 +21.0

Total-U fails below vanilla under pure dynamics shift (70.1% vs. 72.4%) because it incorrectly applies observation recovery to clean sensor readings. Our decomposed approach avoids this cross-contamination.

Uncertainty Trigger Behavior

Trigger rates and cube height trajectory

(a) Aleatoric and epistemic trigger rates under four perturbation conditions. The aleatoric signal fires on 85.2% of steps under sensor perturbation while the epistemic signal fires on only 12.4% under dynamics shift — 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.

Results: Adaptive Visual Tracking

Uncertainty-guided adaptive model selection across YOLOv8 variants (Nano 3.2M to XLarge 68.2M params) on MOT17 (7 sequences, 4,746 frames).

Sequence Frames Nano (%) Small (%) Medium (%) Large (%) XLarge (%) Switches Savings (%)
MOT17-0255018.27.335.332.07.21258.3
MOT17-04105025.24.328.134.77.71857.5
MOT17-0575027.08.029.928.17.01559.5
MOT17-0945026.87.129.732.44.01460.2
MOT17-1059320.615.525.325.313.31857.4
MOT17-1186724.76.927.827.013.61956.2
MOT17-1367523.48.531.229.37.61658.6
Average4,74623.78.229.629.88.61658.2

The policy uses XLarge (the most expensive model) only 8.6% of the time, achieving 58.2% average compute savings while maintaining detection quality within 0.4% IoU of the fixed XLarge baseline.

Adaptive Model Selection Visualization

Adaptive model hopping on MOT17-04

Adaptive model selection for MOT17-04. Top: Temporal uncertainty evolution (epistemic in red, aleatoric in blue). Bottom: Selected YOLO models color-coded by capacity. The policy correlates model scaling with epistemic spikes while ignoring aleatoric elevation, validating orthogonality-aware learning.

Orthogonality Evidence

Joint distribution of aleatoric and epistemic uncertainty

Joint distribution across 21,324 MOT17 detections showing near-zero correlation (r=0.048), confirming the two uncertainties capture distinct mechanisms.

Per-sequence correlation heatmap

Per-sequence correlation heatmap on MOT17. Orthogonality between aleatoric and epistemic signals holds consistently across all sequences (r ≤ 0.05).

Ablation Study

Ablation study comparing decomposed vs total uncertainty

Our orthogonal decomposition achieves 58.2% average computational savings vs. 44.6% for total uncertainty (+13.6% improvement). 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.

Demo Videos

Robotic Manipulation (Isaac Lab)

Franka Emika Panda performing lift-cube task under compound perturbations. The decomposed controller successfully completes the grasp-and-lift despite simultaneous sensor noise and dynamics shift.

Physical Robot + Vision Pipeline

Agilex MiniScout mobile robot with Intel NUC edge platform running adaptive uncertainty-guided model selection for real-time computer vision tasks.

BibTeX

@article{kumar2025uncertainty,
  author    = {Kumar, Divake and Tayebati, Sina and Naik, Devashri and Poggi, Patrick and Rios, Amanda Sofie and Ahuja, Nilesh and Trivedi, Amit Ranjan},
  title     = {TRIAGE: Type-Routed Interventions via Aleatoric-Epistemic Gated Estimation},
  year      = {2025},
  eprint    = {2603.08128},
  archivePrefix = {arXiv},
  primaryClass  = {cs.RO},
  url       = {https://arxiv.org/abs/2603.08128}
}