Glonass : avoid passing corrupt ephemeris and TOW measurements to observables

[vladisslav2011]

Implemented Glonass ephemeris validation (much the same as with BeiDou).

Demonstration: https://www.youtube.com/watch?v=QvpAruQmgy8
GPS+Galileo for comparison: https://www.youtube.com/watch?v=WA-wMrXQyWg

[vladisslav2011]

Have you seen the updated test:

gnss-sdr/src/tests/unit-tests/system-parameters/glonass_gnav_crc_test.cc

Lines 38 to 44 in 10877bd

	for (int k = 8; k < 85; k++)
	{
	std::bitset<GLONASS_GNAV_STRING_BITS> corrupt_bits = bits;
	corrupt_bits[k] = corrupt_bits[k] ? false : true;
	ASSERT_TRUE(gnav_msg.CRC_test(corrupt_bits));
	ASSERT_TRUE(corrupt_bits == bits);
	}

?
I'm checking, that every flipped data bit is being corrected by Glonass_Gnav_Navigation_Message::CRC_test. If not so, the test would fail and the CI would not pass.

It does not give much improvement because single bit errors do not happen frequently (in my tests). The string is either completely broken with many flipped bits and is uncorrectable (ephemeris corruption is handled by the validator in this case) or the string does not have any flipped bits.

[vladisslav2011]

@ZaquL

Maybe to be completly sure, one should check those data?

No problem. Test data added.

[vladisslav2011]

I have all tests passing on my local machine. Let's wait for the CI to complete.

[carlesfernandez]

I’m not sure if this is the way to go. Ephemeris data have their own validation mechanism, and I feel this external validation is somewhat heuristic. In this sense, I prefer the implementation at #667, which seems to stick to the ICD.

However, this is a good idea that worth exploring it further, maybe as an anti-spoofing mechanism. Thanks for your work.

[vladisslav2011]

@ValZapod
It looks like this PR follows additional recommendations from creators of Glonass.

@carlesfernandez
My tests are showing, that existing validation mechanism fails too often with real world signals (a signal from a vehicle-mounted antenna when driving through a city for example). This validator makes GNSS-SDR produce a better track with less position jumps and much more valid points.
I don't think, that this validator may help in case of spoofing. The GNSS-SDR will still switch to spoofed ephemeris after n frames.
Should I rebase the ECC error correction on top of next and open a separate PR?

[carlesfernandez]

Yes, please, make another PR only with the ECC error correction.

[vladisslav2011]

@carlesfernandez
PR ready: #694