Components for positioning system
This device consists of vehicle GNSS module and RFID module as shown in Fig. 1. The former is responsible for the realtime positioning of the RFID reader. It includes GNSS receiving antenna, GNSS chips, GNSS software shown in Fig. 1a. The later module is employed to capture the signals of RFID reader (shown in Fig. 1b) and RFID tags (shown in Fig. 1c) and then to calculate distance.
Location algorithm
Workflow of location algorithm
The principle of proposed algorithm is simulating the radio frequency base station network by rapid movable RFID readers. An RFID tag is equipped on the target points. The RFID reader continues receiving the tag’s signal when the vehicle is moving within the ranging distance, so that distance between RFID reader and tag can be calculated. The location of RFID reader is gotten by the vehicle GNSS. Based on distance and RFID reader coordinate the target tags can be located. This principle is based on the theory of “triangulation” showed in Fig. 2, where A, B and C are three different points for the movable RFID readers respectively, called ranging points; P is the target object which equipped with target tag, called target tag. Figure 3 shows the detailed workflow of algorithm.
The detailed workflow of algorithm includes follow five steps.

i).
Data capture including both GNSS and RFID data when vehicle is moving.

ii).
Data optimization includes filtering bad ranging points data and GNSS data.

iii).
Data synchronization for ranging points and GNSS points based on time.

iv).
Data classification which means the data with specific ID is assigned to the corresponding target tag;

v).
Target tag’ coordinate calculation based on more than 3 ranging points.
Data optimization
In position, some range discontinuity point caused by RFID reader low voltage or some occludes must be removed. The filtered points are used to compose the triangles with target tags according to triangulation, as shown in Fig. 4. When the vehicle is moving, more range points are generated. Every two continuous points jointing one target tag forms a triangle. While vertex angle is less than 15°, this triangle is regarded as illconditioned triangle and removed from the calculation. In Fig. 4a, the angle ∠BP
_{
3
}
C is less than 15°, therefore triangle BP
_{
2
}
P
_{
3
} is to be removed. With this step, only the good ranging points are saved.
Coordinate calculation
As shown in Fig. 4b, the coordinate values of points A, B, C are gotten by vehicle mounted GNSS devices. The distance between two points is calculated, and then the position of target point P can be calculated by equation (1) according to triangulation.
$$ \left\{\begin{array}{c}\hfill {\left({x}_p{x}_a\right)}^2+{\left({y}_p{y}_a\right)}^2={d}_a^2\hfill \\ {}\hfill {\left({x}_p{x}_b\right)}^2+{\left({y}_p{y}_b\right)}^2={d}_b^2\hfill \end{array}\right. $$
(1)
Where, A(x
_{
a
}
, y
_{
a
}) and B(x
_{
b
}
, y
_{
b
}) are two known points, the distance between A and P is d
_{
a
}, and d
_{
b
} is the distance between B and P, and then the P(x
_{
p
}
, y
_{
p
}) is the calculated coordinate. But, based on theory of triangle geometry, there are two triangles if known all three sides of a triangle. It is means there are two possible position of target point P (point P and P
_{
1
}
, as shown in Fig. 4b), which are lies two side of line A and B. From the algebra, equation (1) is a binary quadratic polynomial, and may be has 0, 1 or 2 possible solutions. Integrating the triangle geometry, in this paper, it has two solutions. To determine the right position for target tag, three or more ranging points are included in calculation. Each two of these ranging points jointed target point can form a triangle and get two possible target tag positions. But one of two positions is the same to all possible target tags. For example in Fig. 4b, ranging points A and B can form a triangle with target tag, and get two possible position P and P1. Ranging point B and C can form another triangle and get two candidate positions P and P2. Point P is the common point in two triangles; therefore P is the right target tag position.
However, above is the theory model. In practice, the possible position cannot be same to next solution due to ranging error. The true scene may like Fig. 4c. The two possible target tag position is near. To determine the right position, only two points with smallest distance are used to calculate, which means the higher accuracy. As shown in Fig. 4c, the point P and P
_{
3
} are candidate position. The simple arithmetic mean algorithm is adopted to determine the final position. If the coordinate for P is (x
_{
p
}
, y
_{
p
}), and P_{3} is (x
_{
p3
}
, y
_{
p3
}), the right coordinate of target tag can be calculated by equation (2).
$$ \left\{\begin{array}{c}\hfill {x}_{po}=\frac{x_p+{x}_{p3}}{2}\hfill \\ {}\hfill {y}_{po}=\frac{y_p+{y}_{p3}}{2}\hfill \end{array}\right. $$
(2)