Chasing Ghosts: A Simulation-to-Real Olfactory Navigation Stack with Optional Vision Augmentation

Kordel K. France; Ovidiu Daescu; Latifur Khan; Rohith Peddi

cs.ROFeb 23, 2026

Chasing Ghosts: A Simulation-to-Real Olfactory Navigation Stack with Optional Vision Augmentation

Authors: Kordel K. France, Ovidiu Daescu, Latifur Khan, Rohith Peddi

Organizations: Dept. of Computer Science University of Texas at Dallas Richardson, TX, USA

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.

Figures & tables

Fig. 1: The UAV equipped with the olfactory processing unit (OPU) and sensor harnesses for scent-based navigation. Panel A shows the metal-oxide sensor configuration, and panel B shows the electrochemical sensor configuration.

Fig. 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 leading to the motherboard attached to the belly of the aircraft; (4) the forward-looking time-of-flight sensor for obstacle avoidance; (5) the forward-looking camera; (6) the motherboard and downward-looking time-of-flight infrared sensors attach to the belly (not shown); (7) the two olfactory sensor antennae (MOX sensors shown, but EC sensors lie at the same location when in the proper configuration).

Fig. 3: (A) From top to bottom, this panel shows the altitude, velocity, pitch, and roll from the controller response (blue line) and the command (red line) from the controls algorithms. (B) A diagram showing a partial flight path of the UAV in simulation. (C) From top to bottom, this panel shows the olfactory signal responses for ethanol and nitrogen dioxide.

Fig. 4: Olfaction-Vision Model Architecture. Blue and green boxes construct Component 1 based on COLIP . The yellow boxes construct Component 2 . The output from both models inform reasoning for navigation.

Fig. 5: Illustration of the course developed for olfactory navigation by the UAV. Figure A shows a bird’s-eye-view of the course, where the ground station (laptop icon) is located at the top left and the UAV (quadcopter icon) takes off next to it; Room 2 hosts the target compound and the plume is formed from here. Figure B shows an isometric view of the ideal path the UAV should pursue to locate the source.

Fig. 6: The IMU replay of the UAV traveling over the whole course. Unfiltered olfactory inertial odometry (OIO) [ france_2025_oio_method , france_2025_oio_cal ] along with filtered OIO via EKF appear alongside naked measurements from the IMU. The addition of the olfaction sensors and OPU as payload significantly biased the IMU. This made compound mapping via long-term inertial odometry nearly impossible and drove our decision for the task termination criteria in Section III-H . Measurements are in centimeters.

Method	Sensor Type	Nav. Algorithm	Average Time $\mu_{t}$ (s)	Sigma Time, $\sigma_{t}$ (s)	Best Time, $\beta_{t}$ (s)
Olfaction	MOX	OIO	98.38	14.84	71.59
Olfaction	EC	OIO	112.22	16.70	97.65
Olfaction	MOX	E.SARSA $(\lambda)$	103.99	14.89	82.73
Olfaction	EC	E.SARSA $(\lambda)$	121.85	16.77	97.10
Olfaction + Vision	MOX	OIO	94.03	14.68	66.12
Olfaction + Vision	EC	OIO	107.71	17.07	80.50

TABLE I: Final Results

Aspect	Shigaki et al. (2018/2023 variants)	Burgués et al. (2019)	Duisterhof et al. (2021)	Our Work
Robot Type	Pocket-sized quadcopter (2018) or ground crawler (2023); single unit.	Nano quadcopter (Crazyflie).	Swarm of nano quadcopters (Crazyflie).	Modified DJI Tello UAV; single unit but with custom harness for stereo sensors. Larger payload for dual sensors and vision integration.
Sensors	Single MOX sensor (2018) or dual alcohol + wind sensors (2023).	Single MOX (TGS 8100).	Single MOX per drone + laser rangers.	Dual MOX/EC sensors in stereo (antennae-like); IR for obstacles; vision complementary. First stereo olfaction demo; multiple sensor types for generalization.
Navigation Approach	Bio-inspired (surge-cast/zigzag); real-time with odor frequency.	Predefined sweeping path; bout detection or concentration mapping.	Mapless bug algorithm with PSO; real-time swarming.	Adaptive TD learning (OIO/Q( $\lambda$ )); real-time plume modeling with temporal differences. TD( $\lambda$ ) for plumes; no a priori path; vision for ”last-mile” confirmation.
Course Size	Small ( 1.5-2m start distance; indoor/outdoor open areas).	160m² indoor.	10x10m (100m²) with obstacles.	200m² indoor with rooms. Largest course; more complex (multi-room plume dynamics).
Success Rate	80-90% (2018: non-perfect; 2023: high but variable in complex envs).	N/A (mapping error 1-2m; not framed as success rate).	92% in sim; 11/12 real runs.	100% across 20+ trials. Perfect localization; robust to hysteresis without overfitting.
Real-Time vs. Post-Mapping	Real-time (autonomous onboard).	Post-mapping (builds 3D map after flight; potential for real-time discussed but not implemented).	Real-time (mapless, autonomous).	Real-time (edge-only; no post-analysis). True real-time source localization (not just mapping like Burgués); straight to source.

TABLE II: Comparison to Relevant Prior Works

Fig. 7: We designed the control loop to model the DJI Tello control algorithms and understand how to sequence the olfaction sensors with the other sensors.

Parameter	Definition
$G_{plant}$	Controller
$y_{c}$	Closed loop command
$y$	Current system response
$\cdot{y}$	Rate feedback
$dt$	timestep, in seconds
$k_{i}$	integral gain

TABLE III: PI Controller with Rate Feedback

Parameter	Value
$k_{p}$ (roll)	0.010
$k_{i}$ (roll)	0.002
$k_{d}$ (roll)	0.006
$k_{p}$ (altitude)	0.070
$k_{i}$ (altitude)	0.015
$k_{d}$ (altitude)	0.060-0.080

TABLE IV: PID gains for each PIR controller

Fig. 8: Prediction of the full 6-second diffusion current for chronoamperometry over various time periods at a 100-Hz measurement frequency.

Fig. 9: Prediction of the full 6-second diffusion current for chronoamperometry over various time periods at a 10-Hz measurement frequency.

Parameter	Value
layers	12
layer types	Fully connected only
input dimension	138
output dimension	512
batch size	16
learning rate	1e-4

TABLE V: Olfaction Encoder

Parameter	Value
layers	4
layer types	Fully connected only
input dimension	512
output dimension	512
embedding dimension	512
batch size	16

TABLE VI: FC Graph Associator

Parameter	Value
layers	4
layer types	2 fully connected, 2 GAT Convolution
input dimension	512
output dimension	512
embedding dimension	512
batch size	16

TABLE VII: GAT Graph Associator

Fig. 10: Training losses for both olfaction-vision models. Note that the GAT-based model is more accurate, but requires more compute and inference time, and thus violates our constraints for full edge computation. As a result, we use the MLP-based model for our final results.

-	Off-Plume	On-Plume (Low Concentration)	On-Plume (High Concentration)
Left Sensor $<$ Right Sensor	0	3	6
Left Sensor == Right Sensor	1	4	7
Left Sensor $>$ Right Sensor	2	5	8

TABLE VIII: State Table for Each Navigation Agent

Parameter	Value
number of states	9
number of actions	7
episodes	10000
$\gamma$ (discount factor)	0.9
$\alpha$ (learning rate)	1e-4
$\epsilon$ (greediness)	1.0

TABLE IX: Q(

\lambda

) & Expected SARSA(

\lambda

) Configuration

Fig. 11: Training losses for both Q( $\lambda$ ) and Expected SARSA( $\lambda$ ) algorithms with no blanks modeled.

Fig. 12: Training losses for both Q( $\lambda$ ) and Expected SARSA( $\lambda$ ) algorithms with blanks modeled.

Fig. 13: $Q$ -values of the final reinforcement learning policy.

Parameter	Value	Units	Description
Diffusion	1.0	Dimensionless	Rate of gas dispersion per timestep; conditional on compound
Sparsity	0.0	Dimensionless	Useful for modeling blanks over long distances; conditional on compound
Temperature	20.0	Degrees Celsius	Assume constant air temperature due to short durations and distances
Relative Humidity	50.0	Percentage	Assume constant relative humidity due to short durations and distances
Air Density	1.225	$kg/m^{3}$	Assume constant air density due to short durations and distances
Wind Speed	1.0	$m/s$	Average windspeed along $x$ -direction

TABLE X: Plume Environment Parameters

Fig. 14: An image of the breakout board we designed to down-select the olfaction sensors for ethanol detection.

Fig. 15: Electrical schematic of the breakout board above. The code repository associated with this manuscript provides Gerber files and full schematics for this board.

Fig. 16: One can additively manufacture the full UAV modification kit for both electrochemical and metal oxide sensing configurations on a single 200mm x 200mm print bed.

Robot Navigation

Robot navigation studies how robots perceive, localize, plan, and move safely toward goals in static or changing environments.

See all 554 papers

Scensory: Real-Time Robotic Olfactory Perception for Joint Identification and Source Localization
Sep 11, 2025Yanbaihui Liu, Erica Babusci, Claudia K. Gunsch +1Robot Navigation Robot Localization
Department of Mechanical Engineering and Materials Science, Duke University, Durham, NC, USA. · Department of Civil and Environmental Engineering, Duke University, Durham, NC, USA. · Department of Electrical and Computer Engineering,Duke University, Durham, NC, USA. +1
Multi-Agent Reinforcement Learning for UAV-Based Chemical Plume Source Localization
Mar 12, 2026Zhirun Li, Derek Hollenbeck, Ruikun Wu +4Multi-Agent Reinforcement Learning UAV Navigation
Department of Electrical and Computer Engineering, University of New Mexico, Albuquerque, NM 87131 USA · Department of Mechanical Engineering, University of California, Merced, CA 95343 USA · Department of Electrical Engineering, Colorado School of Mines, Golden, CO 80401 USA +3
Borrowed Eyes: Markerless Nano-UAV Flight with an Active Quadruped Observer
Oct 7, 2026Alejandro Lorite Mora, Dimitrios Arapis, Andrés FaíñaUAV Navigation GNSS-Denied Localization
IT University of Copenhagen, Denmark · Novo Nordisk A/S, Hillerød, Denmark

Sep 11, 2025J/K move · Enter open · S save

Scensory: Real-Time Robotic Olfactory Perception for Joint Identification and Source Localization

Yanbaihui Liu, Erica Babusci, Claudia K. Gunsch, Boyuan Chen

Department of Mechanical Engineering and Materials Science, Duke University, Durham, NC, USA. · Department of Civil and Environmental Engineering, Duke University, Durham, NC, USA. · Department of Electrical and Computer Engineering,Duke University, Durham, NC, USA.+1

Robot Navigation Robot Localization Robotic Perception

Open paper AskPDF arXiv

Olfaction offers robots access to chemical information that is largely inaccessible to vision, touch, and audition, yet using airborne chemical signals for spatial perception remains challenging because local volatile organic compound (VOC) measurements are shaped by complex chemical transport and sensor dynamics. We introduce Scensory, a robotic olfaction framework that learns to jointly infer biological source identity and relative location from short temporal VOC measurements. Using a robot-automated data collection platform, we pair VOC dynamics from cross-sensitive gas sensor arrays with spatial supervision and train models to predict fungal identity, source direction, and distance. We show that a single sensor array can extract all three quantities from only 3 s of local measurements under ambient environmental conditions, achieving species classification accuracy of up to 80.13%, directional accuracy of up to 68.65%, and mean absolute distance errors of 0.110-0.131 m. Incorporating measurements from multiple spatial locations further reduces ambiguity, improving peak species and directional accuracies by 9.72 and 18.66 percentage points, respectively. We then embody this learned olfactory perception on a mobile robot, where successive local predictions acquired during motion are transformed into a world-frame evidence map, allowing observations from different positions and headings to reinforce persistent source hypotheses and guide closed-loop localization. Across eight selected indoor runs, the robot achieves a planar endpoint error of 0.606 +/- 0.294 m. Our results establish airborne chemical dynamics as a viable perceptual signal for robots to recognize biological sources, reason about their spatial origin, and autonomously navigate toward them under ambient environments.

Abstract

Robot Navigation

Explore similar work

Scensory: Real-Time Robotic Olfactory Perception for Joint Identification and Source Localization

Multi-Agent Reinforcement Learning for UAV-Based Chemical Plume Source Localization

Borrowed Eyes: Markerless Nano-UAV Flight with an Active Quadruped Observer