Figure 1. GPS Trimble 5700

Marking the position of any suspect object is mandatory in demining.  In an automated or semi-automated system, a computer database would seem to be the most efficient way to keep a record of alarms.  First, however, the potential alarms must be located accurately.  GPS technique is a good candidate for this task, as it is simple to use and accurate enough. 

The DYLEMA project’s requirements state that alarms must be located with an accuracy of about ±2 centimetres.  This accuracy can be reached with the real-time kinematic (RTK) technique using  an additional GPS antenna placed at the operator station.  With these preliminary specifications the DGPS 5700, manufactured by TRIMBLE, was selected for our application (see Figures 1 and 2).

Despite of the advantages of DGPS techniques for locating mobile robots outdoors, the degradation or loss of the satellite signal can cause errors in position accuracy. Thus, complementary positioning systems such as odometry or dead-reckoning techniques are sometimes required.

Both DGPS and dead reckoning have some shortcomings, but the final result can be dramatically improved by merging the two techniques.  The merging method is termed sensor fusion, a method allowing for the integration and fusion of data obtained from different sources to produce more accurate information.

In this particular case, we use the Kalman filter, a well-known and easy-to-apply set of mathematical equations for estimating the state of a process. The process state in our localization problem is defined by the x and y positions and orientation, α, of the robot. The z component of the robot position does not provide information because we know that potential alarms are on or underneath the ground surface in the vertical of the detection position.

The localization algorithm is a classical extended-Kalman-filter problem in which the estimation phase is performed using joint position and compass measurements and the update phase employs the DGPS data (see Figure 2).

Figure 2. Locator components