I have never heard of any practical single-antenna GNSS-based attitude determination systems. All known examples employ either IMU aiding, or multi-antenna configurations.

Moreover, I cannot suggest any reliable physical principle on which such a system can be based. In terms of SNR, a GNSS antenna can be sensitive to the incidence angle only if it has a very asymmetrical gain pattern, but this would be in a bad contradiction with general design goals, according to which an ideal GNSS antenna must have a perfectly hemispherical gain pattern.      

The constraints of a L1 receiver to obtain centimeter accuracy

Kinematic surveys centimeter, commonly called "Stop and Go" or "OTF" with L1 GNSS-receivers require a little more attention. It must be understood that with a single frequency, certain constraints apply to achieve centimeter accuracy. However, with the right equipment, several use cases lend themselves perfectly.

There are mainly two important issues:

1- The capability of GNSS module detect (using its firmware) discontinuity in the measurements of the carrier phase, commonly called "cycle skipping";

2- GNSS observations of continuity from one era to another on at least 5 satellites for a minimum period of 30 to 40 minutes, required to accumulate enough data to properly resolve the phase ambiguities.

Typically, data is recorded at a second interval in the field. In this range, it is virtually impossible to detect cycle slips in motion with a resolution of one cycle. It should normally fall back on the ability of GNSS module provide that information each time. Firmware normally have access to a very high sampling rate that allows them to detect cycle slips on the signal from each satellite and to include this information in the observations recorded by the data collection software (can the detect, but can not correct).

The second point is crucial in order to achieve centimeter accuracy. The discontinuity in the comments generally comes from obstructions around the antenna, which is blocking the signal and creates many cycle slips. With a GNSS-L1 receiver, avoid as much as possible the highly obstructed environments. To ensure centimeter accuracy, the records must contain at least 30 periods of 40 minutes straight without interruption for at least 5 satellites. If working with a post-processing software that enables the use of known coordinates (during the survey) to initialize the phase ambiguities, this period can easily be reduced. Also, if the field data collection software can detect signal interruptions, the user can easily go on to reset a point near to revive its statement centimeter.

In the end, we understand that the L1 GNSS-receivers (properly addressing the two issues described earlier) used in kinematic mode must be used in relatively open terrain. In such cases, these devices, used in conjunction with a good antenna and good post-processing software, will give results equivalent to those obtained with GNSS receivers L1 / L2 expensive.

The methodology is to first consider the benefits of using a GNSS multi-antenna receiver with the ability to handle all types of GNSS signals from different antennas (eg. 2-8). The diversity and frequency of the GNSS signals, advanced processing of multiple paths and new algorithms highly sensitive signals will be studied using a universal architecture of a GNSS receiver to open and patented heart. A second approach will be based on principles of cognitive radio (CR) in telecommunications. The proposed integrated receiver will be called multi-antenna GNSS receiver cognitive (MACGR)

Dear Fariz, please read my question once again. I did not ask, what you have elaborated!

I have never heard of any practical single-antenna GNSS-based attitude determination systems. All known examples employ either IMU aiding, or multi-antenna configurations.

Moreover, I cannot suggest any reliable physical principle on which such a system can be based. In terms of SNR, a GNSS antenna can be sensitive to the incidence angle only if it has a very asymmetrical gain pattern, but this would be in a bad contradiction with general design goals, according to which an ideal GNSS antenna must have a perfectly hemispherical gain pattern.      

You must have more than one GNSS reception antenna (generally 4)  for the attitude determination, but with one receiver, using pseudo range measurements, the accuracy is not very useful, and for differential positioning, with more accuracy (cm) you must use phase measurements, and more than one receiver