Hide and Seek is a project based on the paper "Distributed Data Fusion" by Mark Campbell and Nisar Ahmed. The purpose of this project is to show how information can be shared among a network of agents to improve the state estimate of all agents on the network. To explore this, a simple example based on a game of hide and seek that continues to get more elaborate. Currently, the following examples are available:
- Static Process, Stationary Linear Sensors
The project is written in python3 and the following instructions are for Windows. For MAC or Linux, use the operating specific language in place of Windows commands. To install, download the ZIP from github and unpack it on your computer.
This project is dependent on Matplotlib, NumPy, and SciPy. To install the required dependencies, open a terminal, navigate to the directory where the project was saved, and install requirements.txt by typing the following:
$ pip install -r requirements.txt
The project is currently run from the command line. To implement and example, open a terminal, navigate to the location where the project is saved, and enter the following:
$ python program.py
Then follow the prompts to view the desired example. The following is a snapshot of the animation of:
$ This program runs the following exercises:
$ [1]: Static Process, Stationary Linear Sensors
$
$ Select an exercise would you like to run: 1
The seekers' positions on the left side of the map are marked by x's. The hider's position is marked by a red x on the right side of the screen. The seekers are receiving noisy, linear 2-D position measurements from the hider, which is not particularly realistic since most hiders wouldn't want to knowingly broadcast their position to those looking for them. However, this example serves as a straight forward starting point to implement the channel filter. The noise on the sensors was modeled as independent Additive White Gaussian Noise on both the position measurements for x and y. Seeker 1 has a covariance defined by diag([250 m^2, 250 m^2]) while the remaining seekers have covariances defined by diag([500 m^2, 500 m^2]). This means that Seeker 1 has a significantly more reliable sensor compared to the rest of the seekers, though none of the sensors are particularly good.
For this example, Seeker 1, pictured at the top left of the map at (-40, 40), is estimating the state of the hider at (30, 0) independently from the other seekers. The Control Position located at (-25, 5) is a Pseudo-Seeker that is meant to directly compare the results from Seeker 3 at (-40, 0). This Pseudo-Seeker uses the exact same measurement generated by Seeker 3, but does not have the benefit of sharing information with the network. The remaining seekers, Seeker 2 through Seeker 5, share information according to the dashed blue lined that connects them (ie. Seeker 2 (-20, 20) can only communicate with Seeker 3, Seeker 5 (-40, -40) can only communicate with Seeker 4 (-20, -5), whereas Seeker 3 communicates with Seeker 2 and Seeker 4).
The color coordinated diamonds on the right side of the screen represent the 2-D state estimate from the seeker with the corresponding color. Additionally, the ellipse with the same color represents the associated 2-D, two sigma (95%) error bound. This means the seeker is 95% confident the true position of the hider is within this ellipse. Individual measurements are plotted as small squares for each time step. In the example figure, the trends are as expected, with the networked seekers' estimates being extremely close to the true position with error bounds much smaller than those of the independent seekers (Seeker 1 and the Pseudo-Seeker). Additionally, the independent seeker outperforms the Pseudo-Seeker as expected due to the independent seeker having a significantly better sensor.
Use the GitHub issue tracker for:
- Bug reports
- Documentation issues
- Feature requests
This project operates under the MIT License. For more information see LICENSE.
Jack Center