A Fleet of Artificial Chemosensing Moths for Distributed Environmental Monitoring

       

(a) Figure showing the UAV to be used as the basis for this project (b) In this robot experiment a neuronal model of the sensorymotor pathway underlying moth chemotaxis was evaluated. The environment consisted of a 0.7 x 0.5 m. transparent plate onto which a moving grating was projected. The brightness of the image, sensed with a robot mounted video camera, represents the concentration of the pheromone while an infra-red source defined the wind direction. At the bottom of the figure the trace (magenta) of the averaged responses of the simulated ORCs sensitive to the pheromone is displayed. The trajectory of the robot (red) shows that in the presence of the pheromone the motor control circuit for surge dominates the behavior, while in the absence of the pheromone the system displays casting driven by the simulated flip-flop neurons. These preliminary experiments showed that a moth based model of chemotaxis can reliably control a robot in real-time.

 

Project Overview


We will develop a novel technology for efficiently identifying, localising, and mapping complex chemical information within uncertain and realistic chemosensory environments. Unmanned Aerial Vehicle (UAV) technology can be applied to this problem providing adequate sensor and control technology. Our approach will exploit the collective behaviour of autonomous moth-based Chemosensory UAVs (cUAVs), the sensory and control information processing subsystems of which will be based exclusively on models of information processing in insect olfaction. Each individual cUAV will be capable of autonomous behaviour including exploration, obstacle avoidance, and foraging. The cUAV will be equipped with chemical and visual sensors and will autonomously navigate, react to environmental stimuli, and assess the chemical composition of its environment. This development continues our research in artificial and biological olfaction, sensory processing and analysis, neuronal models of learning, real-time behavioural control, and robotics. These cUAV artefacts will be co-ordinated for robust exploration, search, and identification behaviour based upon chemical cues. The effective mapping of the chemosensory environment will be achieved through the collective behaviour of a fleet of these agents co-ordinated from a centralised ground station. Fleets of cUAVs will be deployed to sense and map the airborne chemical composition of large-scale environments. Although we will demonstrate this approach for environmental monitoring, as a combined system for highly robust and efficient chemical/biochemical exploration and localisation this technology also has enormous potential for application in

High level objectives for this project are:

  1. Develop a Chemosensory UAV that uses onboard chemical and visual sensors to autonomously navigate outdoors. The cUAV's mission is to identify volatile compounds and locate their sources.
  2. Map the chemical composition of the environment using a new class of chemical sensors and information processing technologies designed for:
    a. Measurement of chemical concentration,
    b. Classification of chemical composition,
    c. Automatic sensor recalibration.
  3. Implement mechanisms and models of adaptive sensory classification, sensory-motor integration, and action selection. These technologies are based on our investigation of insect strategies of sensory processing and control and their application to robots.
  4. Deploy a fleet of cUAVs to collectively solve the task of mapping a chemosensory environment. The main components of our cUAV will be chemical sensor arrays complete with a wide range of broadly tuned chemosensors (supplied by Alpha MOS SA, France) adapted from a separate EU RTD project, antennal lobe model for encoding the chemosensory stimulus (University of Leicester, UK), distributed adaptive control (DAC) subsystem, motor subsystem, visual system, and mechatronics to drive the device (ETH, Switzerland).

Specific objectives relating to the work programme are:

  1. Build a Chemo-sensing cUAV: develop an cUAV that uses on-board chemical and visual sensors to autonomously navigate outdoors. The cUAV's mission is to identify volatile compounds and locate their sources in complex indoor and outdoor environments.
  2. Odour Based Navigation: demonstrate an ability to conduct chemotaxis behaviour in steady-state odour concentration gradients and complex turbulent odour plumes in indoor and outdoor environments.
  3. Learning within a Realistic Chemo-sensory Environment: demonstrate an ability to discriminate between complex odour blends during navigation and learn odour cues as a result of behavioural conditioning.
  4. Insect Based System for Obstacle Avoidance and Visually Guided Navigation: integrate insect based control systems for obstacle avoidance, course stabilization, and terrain following.
  5. Learning of Behavioral Sequences Applied to Active Sampling of Chemosensors: apply for the first time a neuronal model of sequence learning to the cUAV's task in order to learn optimal behavioural patterns for exploration and sampling.
  6. Collective Sampling and Mapping of Chemical Environments: construct a fleet of cUAVs and base station for the efficient and robust mapping of chemical environments.
  7. Sensory Encoding Optimisation with Learning: to achieve in our cUAV an ability to adapt to salient odour stimuli through optimisation of sensory encoding at the level of the antennal lobe.
  8. Fusion of Sensory Data: to achieve in our cUAV navigation by integrating sensory data from multiple modalities. This objective provides the main interface to other projects within the EU-FET Neuroinformatics proactive intiative.
  9. Localised Adaptation to Compensate for Changing Sensor Characteristics: to achieve an ability for compensation within our cUAV to changes in chemical sensor characteristics over time (temporal drift), by adopting a convergent front-end architecture as used in the biological olfactory pathway of the moth.
  10. Odour Intensity and Odour Quality Discrimination: to achieve an ability in our cUAV to separate odour concentration (intensity) and odour quality in real-time. Biological olfactory systems are adept at distinguishing between odour quality and intensity. Such a property would be of great benefit to machine olfaction applications.
  11. Hyperacuity and Sensitivity Enhancement: by understanding principles of hyperacuity in the olfactory pathway, implement a chemosensory system that can demonstrate higher overall system sensitivity to stimuli than provided by the sensitivity of individual sensing elements.

This project has been funded by the EU-IST-FET Programme under the Fifth Framework (AMOTH – IST-2001-33066 -- start date January 2002.)

 

Project Partners:

Tim C. Pearce, University of Leicester, UK (Antennal lobe modelling and project co-ordinator)

Paul F.M.J. Verschure, ETH, Switzerland (UAV development and sequence-based learning)

Bill Hansson, Sveriges Lantbruksuniversitet, Sweden (Electrophysiology and behavioural experimentation)

Alpha MOS, Toulouse, France (Chemical sensor development)

 

Research Team:

Positions Available

 

Log into project pages (coming soon)

 

Results will be posted here as they become available ...

 

Author: Tim Pearce , last updated 18th October 2001.
Any opinions and views expressed are the author's and not those of the University