We evaluated the performance of the GPS subsystem by collecting data while stationary under 3 conditions: in an open field, near buildings, and under heavy foliage. We took the standard deviation of the data and the results are in Table 1.
We then collected GPS data while walking along multiple paths, and applied 3 point, 5 point, and 10 point moving averages with the Kalman filter at the same time while walking. This data was saved in different columns in our data.txt file. A comparison between the moving averages can be found in Fig. 1. and a comparison between the Kalman filter and 5 point moving average can be found in Fig. 2.
We conducted two tests to evaluate the performance of the LoRa subsystem. We placed one node stationary at one end of Forest Park. A team member then walked away from the stationary node through the park, sending GPS data at different distances to test the maximum effective range of the radio. We determined this range based on the distance at which the transmitted messages began to degrade to the point that our interpretation program could no longer extract usable information. A similar test was conducted in an urban environment. We positioned the stationary node in downtown Clayton and walked along Brentwood Blvd.The results of test tests can be seen in Table 2.
| Situation | Standard Deviation of Raw GPS Data (m) | ||
| Trial #1 | Trial #2 | Trial #3 | |
| Open Field (Unobstructed) | 1.3 | 3.1 | 2.6 |
| Near Obstacles (Buildings) | 6.8 | 8.9 | 9.2 |
| Heavy Foliage (Trees) | 4.6 | 3.6 | 5.1 |
Figure 1. 3-point, 5-point, and 10-point moving average of GPS Longitude
Figure 2. Raw GPS data, 5-point moving average, and Kalman filter performance over time.
Figure 3. Kalman Filter performance in an urban environment.
| Situation | Reliable Communication Range (km) | ||
| Open Field (Unobstructed) | 3.1 | ||
| Urban Environments (Buildings) | 1.8 | ||
Analysis & Final Deliverables
The CLEARPath project successfully met its design objectives. The system proved capable of collecting GPS data with an accuracy of ± 1 meter and transmitting this data over distances exceeding 2 kilometers. These achievements directly address the critical needs of humanitarian demining operations for precise, real-time positional data and robust communication across potentially hazardous areas.
In open-field conditions, the Kalman filter provided the best results in filtering the GPS data. One reason for this might be that we were moving at a more constant velocity when out in the open, which was one simplifying assumption in our state-space model. In urban environments and under dense foliage, a moving average was better able to filter out noise and preserve the general trends of GPS. A reason for this might be that in these environments, the team member’s velocity changed constantly, and our Kalman filter could not account for those changes. We recommend using a combined approach moving forward, utilizing the moving average under extremely noisy conditions, and the Kalman filter in open-field conditions.
We delivered a prototype of a completed system capable of accurate location tracking within a 50 x 50 m grid, communication of GPS readings to each transceiver using a LoRa radio system, and compilation of GPS readings to create a path through a defined region. Additionally, to write a Python program that generates polygonal feature layers in ArcGIS Pro.