Location assistance via LwM2M - usability of the lwm2m location assistance lib

[Tjaz]

Greetings,

we are developing a product using the nrf9160 that requires fast and accurate GPS fixes and uses LwM2M for cloud communication and device management. We are using NCS v2.4.2.

Since fast and accurate fixes are needed, we would like to use the location library, which will allow us to use GNSS (with P-GPS and A-GPS) and cell tower positioning. The library can be configured to not use nrf cloud directly, but to let the app handle assistance injection (using CONFIG_LOCATION_SERVICE_EXTERNAL)  - this should allow us to use lwm2m to download the assistance data.

The above is fine and well-documented.

Now, looking at lwm2m, we have the lwm2m location assistance, which is marked as experimental, but has, as far as I can tell, remained unchanged since NCS v2.2.0. 
Looking at the code, I can see the following relevant files:
- nrf/subsys/net/lib/lwm2m_client_utils/lwm2m/gnss_assistance_obj.c
- nrf/subsys/net/lib/lwm2m_client_utils/lwm2m/include/gnss_assistance_obj.h
- nrf/subsys/net/lib/lwm2m_client_utils/location/location_assistance.c
- nrf/include/net/lwm2m_client_utils_location.h

Sorry to be rude, but this is a mess. Just some issues:

- gnss_assistance_obj.c implements functions that are specified in lwm2m_client_utils_location.h (for example location_assist_agps_request_set).
- The function location_event_handler_init from lwm2m_client_utils_location.h is not implemented in any of the above .c files. It actually belongs to the lwm2m sample (nrf/samples/nrf9160/lwm2m_client/src/events/location_event_handler.c). 
- retry_seconds in location_assistance.c is initialized to 0 instead of INITIAL_RETRY_INTERVAL, which causes location_assist_ground_fix_result_cb to schedule the retry work immediately if LOCATION_ASSIST_RESULT_CODE_TEMP_ERR is received before any LOCATION_ASSIST_RESULT_CODE_OK are received.
- gnss_assistance_obj.c uses the nrf_cloud lib directly when receiving a-gps orr p-gps data. For example, in gnss_assist_write_pgps, the function nrf_cloud_pgps_process_update is called. Reading the location lib documentation above, it wants us to call location_pgps_data_process with the p-gps data (which calls nrf_cloud_pgps_process internally). 

I have the following questions:

1. If we want to integrate the location library and the lw2m location assistance, what is the best course of action?
   a) should we use the gnss_assistance_obj.c as is. Will the location lib know that the p-gps data was injected correctly, even if we don't call location_pgps_data_process?
   b) Should we register our own write handlers (gnss_assist_write_cb, gnss_assistance_result_code_cb) and use those to write to the location lib as specified in the location docs?
   c) Use the gnss_assistance_obj.c as a "base" and implement our own, which correctly integrates with the location lib.
   d) something else?

2. What is the timeline on improvements and finalization of lwm2m gnss assistance? It would be nice to use something official form the SDK as is, but the current state of the lib is not usable for us. Are there any short-term plans to improve this? Do you plan on including this as part of the location lib under the hood (similarly to nrf cloud and here positioning)?

Making modifications/using our own implementations of the location assistance objects is not ideal, since they are not part of the lwm2m spec. Thus, they might change as you (Nordic) or AVSystems modifies them. Thus, It would be also nice to know if you plan on replacing these objects (as you have done with object 50001 in the past).

Please advise on the best course of action for us.

Regards,
Tjaž

[Naeem Maroof]

Hello Tjaž,

Thank you for contacting DevZone at NordicSemi.

And thank you for providing comments and related queries. I do not have any information about it, but I have asked internally about your queries and will update you as I receive information. 

with regards,

Naeem

[Tjaz]

Thanks,

is there any timeline for the response?

[Naeem Maroof]

Hello Tjaž,

and thank you for your patience.

Normally we get a response within 2-3 working days, but sometimes some developers might not be available at a time.

Nonetheless, I have received the response now, which I am putting as below:
(Some developers have worked with LwM2M and others with the location services):

Currently we do not have anything at back log. What is missing from current location assistance? The function gnss_assist_write_pgps() handles the P-GPS data from cloud. It will call nrf_cloud_pgps_process_update(). I don´t see anything missing from P-GPS.

Also,
The modem shell has one example of how to handle agnss data with location library (location_srv_ext_agnss_handle):
https://github.com/nrfconnect/sdk-nrf/blob/main/samples/cellular/modem_shell/src/location/location_srv_ext_lwm2m.c

And,
we would like to point to the modem_shell implementation. While being mostly a test tool, modem_shell has support for A-GPS, P-GPS and cellular location using LwM2M and the Location library. Here are instructions how to build modem_shell with LwM2M support: https://developer.nordicsemi.com/nRF_Connect_SDK/doc/latest/nrf/samples/cellular/modem_shell/README.html#lwm2m-support

BR, Naeem

[Tjaz]

Hello,

thank you for the links. The implementation in the modem shell is useful and demonstrates what we were looking for.

Currently we do not have anything at back log. What is missing from current location assistance?

The functionality is all there, but it is undocumented in v2.4.2.
I can see that you have updated the documentation in the latest branch (https://developer.nordicsemi.com/nRF_Connect_SDK/doc/latest/nrf/libraries/networking/lwm2m_location_assistance.html). You updated it yesterday - so saying "there is nothing in the backlog" is a bit disingenuous. You are literally working on it right now. You could have said: "Expect improvements in the next NCS release", or something.

I think you still have a bug here: https://github.com/nrfconnect/sdk-nrf/blob/main/subsys/net/lib/lwm2m_client_utils/location/location_assistance.c#L61

retry_seconds is initialized to 0 (since it is static). It should be initialized to INITIAL_RETRY_INTERVAL. With the current code, if you get a temporary error response from the server, the retry work will be scheduled immediately.

Regards,
Tjaž 

[Naeem Maroof]

Hello Tjaž,

and I am sorry if you felt like that. As I said earlier, I do not know about it and just kind of relayed the response from different developers/teams to you.

Please expect improvements in upcoming NCS releases.

And thank you for point-out and commenting on the potential bug. I will let them know about it.

Kind regards,
Naeem