Research Analyzer
← Back ICRA 2026

A Narwhal-Inspired Sensing-To-Control Framework for Small Fixed-Wing Aircraft

Fengze Xie, Xiaozhou Fan, Jacob Schuster, Yisong Yue, Gharib Morteza

PDF

AI summary

Key figure (auto-extracted from paper)
Combining standoff airflow sensing with physics-informed dynamics learning significantly improves force estimation and control agility for small fixed-wing UAVs in turbulent conditions.
Fixed-wing UAV Sensing-to-control Physics-informed learning Multi-hole probe Control allocation Gust robustness

Problem

Small fixed-wing UAVs struggle with low-speed agility and accurate aerodynamic sensing due to highly coupled dynamics and near-body flow distortions from the airframe and propellers.

Approach

We mount a standoff multi-hole probe and sparse wing pressure sensors to estimate flow state, then learn a symmetry-regularized control-affine dynamics model to map these observations to aerodynamic wrenches for convex control allocation.

Key results

  • Wing pressure sensors reduce force-estimation error by 25%–30%
  • Physics-informed model degrades only 12% under distribution shift versus 44% for unstructured baselines
  • Normal-force tracking RMSE improves by 27% over plain affine models and 34% over unstructured baselines
  • Convex control allocation yields smoother, well-trimmed actuation commands

Why it matters

Provides a lightweight, robust sensing-to-control pipeline that enables safer, more agile fixed-wing UAV operations in real-world turbulent environments without heavy payload penalties.

Abstract

Fixed-wing unmanned aerial vehicles (UAVs) offer endurance and efficiency but lack low-speed agility because its highly-coupled dynamical model. We present an end-to-end sensing-to-control pipeline that combines bio-inspired hardware instrumentation, physics-informed dynamics learning, and con- vex control allocation. Measuring airflow on a small airframe is difficult as near-body aerodynamics, propeller slipstream, con- trol surfaces actuation, and the present of gusts would distort pressure signals. Inspired by the narwhal whale’s signature protruding tusk, we stick our in-house developed multi-hole probes far into the upstream, and complement it with sparse yet carefully placed wing pressure sensors for local flow measure- ment. A data-driven calibration scheme was adopted to map the pressure signal of the probes to airspeed and flow angles. We next learn a control-affine dynamical model using Pitot- tube estimated airspeed and flow angles as inputs, along with sparse sensor measurements. We implement a soft left/right symmetry regularizer that improves the model’s identifiability under partial observability and limits confounding between wing pressures and flaperon inputs. Desired wrenches (forces and moments output) are realized by a regularized least-squares optimizer that yields smooth, trimmed actuation. Wind tunnel studies, across a wide range of parameter space, demonstrate that adding wing pressures reduces force-estimation error by 25%–30%, the proposed model degrades less under distribution shift (about 12% versus 44% for an unstructured baseline), and force tracking improves with smoother inputs, including a 27% reduction in normal-force RMSE relative to a plain affine model and 34% relative to an unstructured baseline.

Index terms

Model Learning for Control Machine Learning for Robot Control Aerial Systems: Mechanics and Control

Related papers