[Paper Review] Chasing Ghosts: A Simulation-to-Real Olfactory Navigation Stack with Optional Vision Augmentation

Kordel K. France, Ovidiu Daescu+2 more|arXiv (Cornell University)|Feb 23, 2026

Insect Pheromone Research and Control0 citations

TL;DR

The paper presents an open-source UAV system that localizes an odor source using minimal olfactory sensors, with an optional vision module, trained in simulation and deployed on a real quadrotor.

ABSTRACT

Autonomous odor source localization remains a challenging problem for aerial robots due to turbulent airflow, sparse and delayed sensory signals, and strict payload and compute constraints. While prior unmanned aerial vehicle (UAV)-based olfaction systems have demonstrated gas distribution mapping or reactive plume tracing, they rely on predefined coverage patterns, external infrastructure, or extensive sensing and coordination. In this work, we present a complete, open-source UAV system for online odor source localization using a minimal sensor suite. The system integrates custom olfaction hardware, onboard sensing, and a learning-based navigation policy trained in simulation and deployed on a real quadrotor. Through our minimal framework, the UAV is able to navigate directly toward an odor source without constructing an explicit gas distribution map or relying on external positioning systems. Vision is incorporated as an optional complementary modality to accelerate navigation under certain conditions. We validate the proposed system through real-world flight experiments in a large indoor environment using an ethanol source, demonstrating consistent source-finding behavior under realistic airflow conditions. The primary contribution of this work is a reproducible system and methodological framework for UAV-based olfactory navigation and source finding under minimal sensing assumptions. We elaborate on our hardware design and open source our UAV firmware, simulation code, olfaction-vision dataset, and circuit board to the community. Code, data, and designs will be made available at https://github.com/KordelFranceTech/ChasingGhosts.

Motivation & Objective

• Demonstrate online odor source localization on a lightweight UAV without explicit gas distribution mapping or external positioning.
• Develop a minimal-sensor navigation framework that works across olfactory sensing modalities.
• Enable sim-to-real transfer by training navigation policies in simulation and deploying them on hardware.
• Incorporate an optional vision modality to accelerate odor-guided navigation under certain conditions.
• Provide reproducible hardware, software, and datasets to foster future olfactory robotics research.

Proposed method

• Integrate custom olfaction hardware (MOX and electrochemical sensors) and infrared sensing on a modified DJI Tello platform.
• Train navigation policies in simulation (Gymnasium-based plume environment and Kalman-filtered control loops) and deploy on hardware with a minimal sensor suite.
• Use dual-timescale odor processing (short/long EMA differences) to detect plumes and compute plume angles for steering.
• Develop COLIP (contrastive olfaction-vision pre-training) with a CLIP-based vision encoder and a 12-layer olfactory encoder to align olfactory and visual embeddings via InfoNCE loss.
• Implement a two-component vision system: a slow, high-level OVM and a fast, YOLOv11-based reactive module to balance generalization and edge-speed inference.
• Translate plume-directed decisions into a discrete action set (surge, cast directions, pause, land) via olfactory inertial odometry and plume dynamics.

Figure 1: The UAV equipped with the olfactory processing unit (OPU) and sensor harnesses for scent-based navigation. The metal-oxide sensor configuration is shown in panel A with the electrochemical sensor configuration shown in panel B.

Experimental results

Research questions

• RQ1 Can a UAV locate an odor source online using only a minimal set of olfactory sensors without prior gas distribution maps or external localization aids?
• RQ2 Does optional vision augmentation improve odor source localization performance and speed, and how does it interact with olfactory-driven control?
• RQ3 How well does sim-to-real transfer work for olfactory navigation policies on a real UAV under realistic airflow conditions?
• RQ4 Is the proposed minimal-sensor framework generalizable across different sensing mediums (MOX vs electrochemical) and environments?

Key findings

• Olfaction-alone navigation successfully localizes the odor source across trials with MOX and EC sensors.
• Adding vision consistently reduces navigation time compared to olfaction-only navigation across sensor types.
• Vision augmentation yields performance gains, with average task times improved relative to olfaction-only baselines.
• Sim-to-real deployment achieved on a real UAV operating over a 200 m² course with real-time olfactory data processing.
• The approach generalizes across sensing mediums and demonstrates reproducibility via open hardware, software, and datasets.

Figure 2: The UAV equipped with the olfactory processing unit (OPU) and sensor harnesses. The figure shows a top-down view of the aircraft: (1) the 2 ballast points needed to balance the aircraft; (2) the battery to power the motherboard required for the olfactory sensors; (3) the wire harnessing le