Dear Alex Minetto ,

Are you looking for datasets where GNSS is collected from multiple receivers located in the surrounding, or also where also some type of inter-agent ranging is present? Probably the availability of the latter is the most challenging to find, when it comes to open dataset!

I am following the question, such a dataset would be really interesting to work with.

Best,

Daniel

Dear Daniel Medina ,

Thanks for participating in the discussion.

I am actually looking for the first (multiple receivers in the sorrounding). I would check if some research group has ever collected similar datasets and make them accessible to be used for our GNSS-only cooperative algorithms.

I am planning to set up a data collection including high-rate PVT solutions of a set of receivers (possibly 15 to 30) in different kinematic conditions (vehicles, pedestrian, static) in a given proximity radius (approx. 5 Km). I would then reproduce the GNSS signal generation through our IFEN NCS Titan thus having all the needed flexibility to synchronize the paths, satellite visibility and signals reception. Our cooperative SwRx will then process the data of each path independently from the others to extract GNSS-based inter-agent range measurements and integrate them in the positioning solution.

Off course, one of the most time-consuming part regard the reference data to be collected but I think there are few high-rate data available in this shape. As anticipated I do not need them to be synchronous but they should be accurate enough to properly describe the dynamics of the motion (limited jerk or higher-order discontinuities would be appreciated).

Best,

Alex

Dear Alex Minetto ,

It would be great if such a dataset exists, although I do not think that is the case.

With regards to your proposal experiment, it sounds really interesting and equally challenging (having 15 to 30 [geodetic] antennas and receivers is on itself a challenge!). After this whole situation is over and if you need some support, let us know whether we can collaborate on your data collection!

Best,

Daniel

Dear Daniel Medina ,

My plan is to collect all the GNSS fixes asynchronously traveling on different trajectories with only one dual-frequency GNSS/INS/RTK high-accuracy receiver mounted on a vehicle. This would provide a realistic set of data (which are very difficult to be obtained analytically withouth proper sw for vehicular simulation, do you have any?).

The recorded accurate state estimation (properly smoothed if needed) will provide the reference of our "accurate" dataset. They must be then independently (track-by-track) replicated off-line in the LAB through a GNSS signal and constellation generator (IFEN Titan).

Also the sampling of the generated RF signals must be performed track-by-track as well as the post-processing through a sw receiver (time consuming but effective). Each run of the sw receiver will estimate the observables with a slight time-misalignment w.r.t. to each other, thus forming a realistic asynchronous scenario in which each receiveir independently process the received signals.

Eventually it will be possible to combine the observables retrieved by a set of receivers observing the same GNSS scenario to test collaborative frameworks and algorithms.

I did the same during my Ph.D. by using MATLAB-generated analytical trajectories instead of realistic vehicular tracks as I would like to do.

What is your feeling about this?

Dear Alex Minetto ,

Sorry for the delayed reply!

I am not sure whether I have understood the "final" aim of the experimentation that you describe above: are you interested in collected simultaneously actual GNSS data across multiple receivers in a nearby location or in having a simulation tool to recreate such scenario?

Also, do you have an interest in keeping the IF samples for the scenario (which would require some massive storage and the adequate GNSS SDR recording HW to do it) or interesting only on the raw observables (code and phase pseudoranges)?

With regards to the tool for the vehicular simulation, the best I can think of consists on inputting a trajectory (with the corresponding date and time stamp) and from this one could discretize the time to generate IMU measurements (again, given the IMU characteristics) and then generate the GNSS observables for such trajectory (this could be made using a constellation generator). This is possible, although I do not know of any open-source SW to do it!

The question remains on whether a simulation tool can assess all the cross-correlations appearing locally among different receivers...

Cheers,

Daniel

Dear Daniel Medina ,

let's say that the aim is in the middle but the actual data collection with multiple high-accuracy receivers is practically infeasible (devices, vehicles, operators). The goal would be hence to get a flexible tool to generate multi-agent scenarios and to provide several products retrieved from their trajectories, to be used at different research layers in GNSS.

• N high-accuracy reference trajectory (ground truth)
• Inertial measurements retrieved from the on-board IMU (for integration purposes)
• N raw samples, binary files (.bin) acquired by front-end from the signal generator (for low-level processing, i.e. sample domain)
• raw measurements sets and related PVT solutions retrieved by the post processing of the recorded signals (for medium and high-level processing, i.e. measurements and solution domains)

Using some analytic trajectories, the workflow was pretty confortable for theoretical analysis during my Ph.D. but real kinematics was far away from the behaviour of my MATLAB-generated virtual vehicles.

A big challange will regard your last sentence. Off course, by generating signals with multiple instances of the signal generator we will observe independent physical processes which is not exactly true for real scenarios. However, the channel between satellites and agents could behave similarly hiding weak cross-correlations phenomena.

Cheers,

Alex