ROBONOSE: Gas Source Localization with a Mobile Robot

1) The Problem

Using electrochemical sensors on a mobile robot is very promising for a broad range of applications. For example, chemical sensing can be useful for an "electronic watchman" to detect, localize and identify existing odours thus indicating leaking solvents, hazardous gases or a fire at its initial stage. Self-produced odours can be also used to aid navigation [Eng89] or to communicate with other robots, for instance by sending a chemical SOS signal [RuKlKe00].

There are, however, a couple of problems with using gas sensors in real world environments. Unlike visual or auditory, chemical stimuli are not inherently directional. Therefore the alignment with respect to a gas source has to be determined from the spatial concentration distribution. But in most indoor environments the spread of gases is dominated by turbulence and weak convection flow [WaLiDu03] rather than diffusion, which is known to be a considerably slower transport mechanism for gases in general [NaIsMo99]. Such concentration profiles show many time-varying discontinous patches of local eddies [Hi75] and do not provide a uniform gradient that can be used to identify the location of the gas source. Remarkably the absolute concentration maximum is usually not located near the gas source if this source has been active for some time. This point is illustrated by Fig. 1, which shows typical sensor readings in the vicinity of a gas source of evaporating liquid ethanol.



Typical Gas Distribution
    
Fig. 1: Typical gas distribution in an environment without a strong uniform airflow.

Approaches for the tasks of detection, localization and declaration of a static gas source has to consider these attributes of gas transport. Another difficulty is posed by the response characteristic of the metal oxide gas sensors used. Due to their slow response and recovery, integration of subsequent readings is always performed implicitly. Fig. 2 shows the dynamic response of a Taguchi-type metal oxide sensor (Figaro TGS 2620) to a step pulse that was generated by opening a bottle of ethanol for 10 seconds in direct vicinity of the sensor.



    
Fig. 2: Gas sensor readings (circles) together with the corresponding response fit (line) that was computed using a first-order sensor model.

2) Concentration Mapping, Gas Source Localization and Declaration

Our research in the field of gas source localisation addresses the following questions:

  • How does the driving strategy of the robot that carries the mobile nose affect its localisation capability ?
    In this concerns it was found that stop-sense-go strategies perform considerably worse than constant-speed-driving [LiWa01a]. This result was confirmed in [MoDu02].
  • Is it possible to locate a static gas source by analysing the response peaks of the gas sensors ?
    While it was found that this is possible if the problem can be reduced to one dimension in certain corridor-like indoor environments [LiWa01a] such a method is usually not appropriate for localisation on a two dimensional base [LiWa01b].
  • To what extent typical environmental conditions can be determined for rooms without a strong uniform airflow ?
    It was found that the concentration profiles in unventilated rooms with a static gas source are usually relatively stable over a period of some hours [WaLiDu03]. This time-invariant structure is of course superimposed by the transient variations created by turbulence. It is likely caused by convection flow due to stationary temperature differences [LiWa01b].
  • How can the concentration profile be extracted from subsequent readings collected along the path of the robot ?
    By contrast to metric gridmaps extracted from sonar or laser range scans, a single measurement of a gas sensor provides information about a comparatively small area (about 1 square centimeter). To overcome this problem an algorithm to build gridmaps was introduced that uses a Gaussian density function to model the decreasing likelihood that a particular reading represents the true concentration with respect to the distance from the point of measurement. With this technique it was possible to represent the time-invariant plume-like structure of a gas distribution in a concentration gridmap (see Fig. 3). The stability of the mapped features with time and the impact of the slow recovery of the gas sensors are discussed in [LiDu03b], while the parameters of the introduced model and the capability of concentration gridmaps to locate a gas source are discussed in [LiDu03c].

Example of a Concentration Map (a)

Fig. 3a: Example of a gas concentration gridmap after 90 minutes of inspectation.

Example of a Concentration Map (b)

Fig. 3b: Example of a gas concentration gridmap after 180 minutes of inspectation.

  • Is it possible to locate a static gas source in a turbulent environment using a purely reactive strategy ?
    To address this question a gas-sensitive Braitenberg vehicle [Bra84] was implemented that realises a direct sensor-motor coupling. The transfer function was chosen to be monotonous and inhibitory. In this way maximum wheel speed results if the sensed concentration is low, which in turn implements a simple sort of exploration behaviour. This behaviour combined with either hillclimbing in case of uncrossed sensor-motor connections or concentration peak avoidance in case of crossed connections. With uncrossed inhibitory connections the average path length the robot needs to move to the source is reduced by a factor of two compared to random search [LiDu03a]. An indirect localisation method results from using crossed inhibitory connections because then the path of the robot covers the available space except near the actual location of the source. This strategy offers also a solution for the task of gas source declaration without using additional sensors. Rather than trying to identify the gas source by a global concentration maximum, the source is recognized by the higher frequency of local concentration maxima in its vicinity [LiDu03a]. To enhance the performance of the suggested indirect localisation method a mapping algorithm was introduced that determines a belief of the location of the gas source by combining single estimates derived from the curvature of the path driven by the robot [Li03].


Fig. 4a: Example of the path driven with uncrossed inhibitory sensor-motor connections (permanent love).


Fig. 4b: Example of the path driven with crossed inhibitory sensor-motor connections (exploring love).

3) Credits

This research is funded by the state of Baden-Württemberg. Furthermore it was supported with a Marie-Curie grant that was part of the European Commission's 5th framework programme (FP5). The particular sub-projects were realised in cooperation with the Institute of Physical Chemistry (ipc) in Tübingen, Germany and the Center for Applied Autonomous Sensor Systems in Örebro, Sweden.

The experiments were performed either with the robot "ARTHUR" (see Fig. 5a), which is based on the model ATRV-JR manufactured by iRobot and with the Koala robot (K-Team) that is shown in Fig. 5b.




Fig. 5a: The mobile robot "ARTHUR"


Fig. 5b: The Koala mobile robot

References

  • [Eng89] J. F. Engelberger, "Robotics in Service", Kogan Page, London, 1989
  • [RuKlKe00] R. A. Russell, L. Kleeman and S. Kennedy, "Using Volatile Chemicals to Help Locate Targets in Complex Environments" Proceedings of the Australian Conference on Robotics and Automation, pp. 87-91, 2000.
  • [WaLiDu03] M. R. Wandel, A. J. Lilienthal, T. Duckett, U. Weimar and A. Zell, "Gas Distribution in Unventilated Indoor Environments Inspected by a Mobile Robot" submitted to ICAR 2003, 2003.
  • [NaIsMo99] T. Nakamoto and H. Ishida and T. Moriizumi, "A Sensing System for Odor Plumes", Analytical Chem. News & Features, vol. 1, pp. 531-537, 1999.
  • [Hi75] J. O. Hinze, "Turbulence", McGraw-Hill, New York, 1975
  • [LiWa01a] A. J. Lilienthal, M. R. Wandel, U. Weimar and A. Zell, "Sensing Odour Sources in Indoor Environments Without a Constant Airflow by a Mobile Robot" Proceedings of the IEEE International Conference on Robotics and Automation (ICRA 2001), pp. 4005-4010, 2001.
  • [MoDu02] A. M. Farah and T. Duckett, "Reactive Localisation of an Odour Source by a Learning Mobile Robot" Proceedings of the Second Swedish Workshop on Autonomous Robotics, pp. 29-38, 2002.
  • [LiWa01a] A. J. Lilienthal, M. R. Wandel, U. Weimar and A. Zell, "Experiences Using Gas Sensors on an Autonomous Mobile Robot" Proceedings of EUROBOT 2001, 4th European Workshop on Advanced Mobile Robots, pp. 1-8, 2001.
  • [Brai75] V. Braitenberg, "Vehicles: Experiments in Synthetic Psychology" MIT Press/Bradford Books, 1984
  • [Li03] A. J. Lilienthal, "An Approach to Gas Source Localisation and Declaration by Pure Chemo-Tropotaxis" submitted to ECAL 2003, 2003.
  • [LiDu03a] A. J. Lilienthal and T. Duckett, "Experimental Analysis of Smelling Braitenberg Vehicles" Proceedings of the IEEE International Conference on Advanced Robotics (ICAR 2003), 2003.
  • [LiDu03b] A. J. Lilienthal and T. Duckett, "Creating Gas Concentration Gridmaps with a Mobile Robot" submitted to IROS 2003, 2003.
  • [LiDu03c] A. J. Lilienthal and T. Duckett, "Gas Source Localisation by Constructing Concentration Gridmaps with a Mobile Robot" submitted to ECMR 2003, 2003.


Contact

Achim Lilienthal , Tel.: (07071)29-78988, lilien@informatik.uni-tuebingen.de