NightHawk: Active Illumination Control

ISER 2025

Yash Turkar, Youngjin Kim, Karthik Dantu
Center for Embodied Autonomy and Robotics (CEAR)
University at Buffalo
Arxiv Code Culvert Inspection

Active illumination and exposure control on Boston Dynamics Spot, inspecting culverts under the Erie Canal in Medina, NY. (Left) Spot approaches entrance of culvert 110, a 66m long, subterranean environment with extremely low ambient illumination. (Right) Spot entering and exiting the culvert equipped with NightHawk

Abstract

Subterranean environments such as culverts present significant challenges to robot vision due to dim lighting and lack of distinctive features. Although onboard illumination can help, it introduces issues such as specular reflections,overexposure,and increased power consumption. We propose NightHawk, a framework that combines active illumination with exposure control to optimize image quality in these settings. NightHawk formulates an online Bayesian optimization problem to determine the best light intensity and exposure-time for a given scene. We propose a novel feature detector-based metric to quantify image utility and use it as the cost function for the optimizer. We built NightHawk as an event-triggered recursive optimization pipeline and deployed it on a legged robot navigating a culvert beneath the Erie Canal. Results from field experiments demonstrate improvements in feature detection and matching by 47-197% enabling more reliable visual estimation in challenging lighting conditions.

Coming Soon!

Method

The process begins with an initial optimization to obtain optimal settings (∆t*,P*) and the corresponding metric (M*feat), which are executed by the vision system. The resulting image is evaluated using Mfeat; if the deviation exceeds the threshold (ε), optimization is triggered again; otherwise, the previous settings are reused.

Feature-based Image Utility (Mfeat)

Correlation of utility metrics with feature detection/matching. Mfeat demonstrates a strong positive correlation with feature matching performance across five diverse feature detectors (AKAZE, SHI_TOMASI, ORB, R2D2, and Superpoint).

NightHawk in the Field

Feature tracking performance of the 3 settings. Mean exposure-times (∆t) for each run in top-right corner. AE + no external light (left) shows feature tracking loss when moving through low-light environments while AE + fixed (100%) external light (middle) shows some improvement but is limited due to over-exposure and specular reflections. NightHawk (right) shows overall longest feature tracks, fewest discontinuities and lowest average exposure-times
The change in Mfeat as the robot enters the culvert is shown under two conditions: with auto-exposure and no external light (top), with external light at maximum intensity (center) and with NightHawk (bottom). In the first scenario, the exposure eventually reaches its lower limit, resulting in an under-exposed image that is unusable. In contrast, when external light at maximum intensity is applied, artifacts such as reflections and a greenish hue appear at both the entrance and exit of the culvert. Finally, with NightHawk, Overall image utility is consistent, the sudden dips and green-ish hue at culvert entrance is not seen anymore

BibTeX

@inproceedings{turkar2025nighthawk,
  title     = {Active Illumination Control in Low-Light Environments},
  author    = {Turkar, Yash and Kim, Youngjin and Dantu, Karthik},
  booktitle = {International Symposium on Experimental Robotics (ISER)},
  year      = {2025},
  note      = {To appear. Preprint available at \url{https://arxiv.org/abs/2506.06394}}
}