I had written a simple simulation for trajectory in radar. You could use it simply or change it to whatever you need

I had written a simple simulation for trajectory in radar. You could use it simply or change it to whatever you need

Why dont you use wiimote data:

http://idav.ucdavis.edu/~okreylos/ResDev/Wiimote/

Hello. With a colleague, I recently published a benchmark data set for 1-D passive sonar tracking. This pair of data sets (bearing-time and frequency-time) is challenging since the false alarm density is very high compared with radar data, which typically has 2 or 3-D measurements and therefore a greatly reduced effective clutter density. Hence the passive sonar problem provides a better test environment for MTT algorirhms. Please refer to the following references on data association and to retrieve the benchmark tracking data.Note that this data is for a single-sensor multi-target scenario. The description document provides details for you to repeat the experiment and add other sensors or platforms (or ownship in sonar terminology). Please refer to the benchmark document below if you publish results on this data set. Good luck.

Technical Report Benchmark 4-target passive sonar scenario description for 1-D tracking

Article Taxonomy of multiple target tracking methods

Hello. This response is to help someone who contacted me regarding how to use the data sets that I posted previously. The Matlab code below (read_file.m, 17/3/2015, 72 lines) will allow you to open and read the tracking data (BearingEvents.txt or FrequencyEvents.txt) and run your own tracking/data association algorithms in the loop. Note that you will need to add in your own track initiation, which can be manual (tell it where to start) or automatic (e.g. 2-point initation). Note also that this code is for reading ASCII files with scans separated by a header line of zeros.

If you want to generate data for Monte Carlo simulation or change the scenario, I am afraid that I cannot help with this, but the instructions in the document linked to above should be sufficiently detailed to allow researchers to repeat this experiment.

The value of these data sets is that they have already been used for tracking performance characterisation and are challenging data sets. Both the strongest neighbour and nearest neighbour Kalman filter cannot track either the 1-D bearing or frequency data without multiple track breaks and either many false tracks or missing some of the starts of tracks. You need a good tracking algorithm for this data. Remember that although this data is 1-D, it cannot be treated as an image processing problem since in practice, you only get 1 scan at a time and you cannot wait for the entire data set to arrive!

In practice, it is quite a lot of work to put together software that generates realistic tracking data (such as the data files in the link), which explains why there are limited public domain resources available for free. For most problems, you can hard-code a scenario and just add false alarms and this should be adequate for research purposes, however, you cannot be sure you have results that are directly comparable to those of other researchers unless you use a benchmark data set.