Research on tracking strategy of double-row sensors

At present, most smart cars use a single-row sensor road detection method. This method obtains less road information and cannot distinguish the state of the smart car from the road conditions, causing control problems. In order to make up for the shortcomings, a long-forward single-row sensor road detection method has been developed. This method has a longer detection distance and can determine the direction of the road earlier. To a certain extent, it makes up for the disadvantage of low detection accuracy, but it cannot effectively distinguish the state of the smart car from the road conditions.

The car models in the competition can use cameras or sensors to detect road information. Our car models use double-row infrared tracking. The large forward-looking double-row sensors can obtain more track information, take strategic processing earlier, and form a better driving track. It is an alternative to using complex camera solutions.

It can achieve stable control and smooth acceleration on straight roads; it can move forward in a small curve in an S-curve to reduce the route and the number of steering adjustments. It can achieve early turning and inner turning effects in large bends. Especially in terms of turning, the front and rear rows can jointly predict the bends to extend the physical recognition distance, so as to make early actions and reduce the negative impact caused by the short detection distance to achieve the above effects.

Sensor array layout

In Figure 1, only the receiving tube is used to illustrate the sensor position.

Layout Description

The front row sensors extend far, so when the center of the car deviates from the black line, a larger offset will be generated on the front row sensors. The rear
row sensors extend close, so when the center of the car deviates from the black line, a smaller offset will be generated on the rear row sensors.
The different sensitivities of the front and rear row sensors to the car offset are used to control the car.
In order to make the front and rear rows have a clearer division of labor and collect information from farther away, we tilt the front row sensors at an angle of about 45o, so that the front row has a larger forward-looking distance and can better reflect the advantages and characteristics of the front row.
Straight road recognition method and control strategy

Straight road identification method

(1) When using this method to arrange double rows of infrared sensors, there are five physical methods for distinguishing straight lines. The timing for applying each method is listed in the table below.

The first straight road situation (Figure 2)

After a large left turn, the most likely combination of the front and rear sensors detecting the black line when exiting the turn. Applicable to 90° and 180° left turns. The exit information is obtained in advance, the steering gear turns to the left at a small angle, and the acceleration action is taken at this time to make up for the lack of foresight. When this situation occurs at the S-bend of the track, it does not meet the second recognition method of the straight road, so it will not accelerate.
The second straight road situation (Figure 3)

This situation is a reconfirmation of the first situation. After turning left and passing the first situation, this situation can confirm that there is a straight road ahead and continue to improve the acceleration ability of the car. The control program switches from the curve program to the straight line stability program.

The third straight road situation (Figure 4)

At this time, straight line stability control is adopted. Since the first two situations have been clearly identified as straight roads, this situation only increases the success rate of straight road identification.

The fourth straight road situation (Figure 5)

Similar to the second case, the fifth case is reconfirmed. After turning right and passing the fifth case, this case is experienced again to confirm that there is a straight road ahead, and the acceleration ability of the car is further improved. The control program is switched from the curve program to the straight line stability program.

The fifth straight road situation (Figure 6)

After a big right turn, the most likely combination of the front and rear sensors detecting a black line when exiting the turn. Applicable to 90o and 180o right turns. The exit information is obtained in advance, the servo turns to the right at a small angle, and acceleration is taken at this time to make up for the lack of foresight. When the track's S-curve appears, the second recognition method of the straight road is not satisfied, so acceleration will not be performed.
(2) Straight road recognition, program-assisted confirmation

After entering the curve, the car will oscillate as it moves, so the above five conditions may not be met when exiting the curve. In order to improve the success rate of straight road recognition, a second straight road identification method is added. Both methods work at the same time. After the first method is met, it takes a maximum of 15ms to confirm that it is a straight road.

The program is executed in a loop, and our program execution frequency is 2KHz. Using the timed interrupt (15ms) method, three counters are used to count the three sensors in the middle of the front row (numbered 3, 4, and 5). Each time the program is executed, if one of them detects a black line, the corresponding counter is increased by 1. After calculation, the maximum value that can be counted within 15ms is 31. We set the maximum value of the count. If the required count value is reached within 15ms, it is considered a straight road, and the straight road program is switched and the counter is cleared; if the required count value is not reached within 15ms, the counter is cleared and counted again. For example, the car has a speed of 2m/s and travels 3cm. We only need to judge that it is a straight road within 2~2.5cm. Therefore, the maximum count value is set to 20~25, which is considered a straight road, and the curve program is jumped out.

Of course, a more rigorous method can also be used to judge, just adjust the time and count value of the timing interruption. This condition can always be met after entering the straight road, so as a supplement to the first straight road judgment method, it ensures stable and reliable recognition of the straight road.

Straight-line stability control strategy

After the car turns, the servo will not react quickly enough, so the car will oscillate before it can stabilize. To reduce the oscillation as soon as possible, the following method is used to control the car's movements after turning:

Set the flag in the curve strategy, enter the straight line program, identify the flag, and take correction settings for the formula that controls the steering of the servo. The formula is: q=K1q1+K2q2; where q is the control amount finally sent to the servo, q1 is the return angle value of the front photoelectric sensor, and q2 is the return angle value of the rear infrared sensor. K1 and K2 are the weighted proportional values ​​of the front and rear sensors respectively. Normally, K1 and K2 are 1, and the values ​​are changed when necessary.

When the car enters the straight road from the curve and successfully identifies the straight road, the value of K1 is reduced. Since the rear sensor is very close to the front wheel (steering wheel) of the car, when the center of the car deviates from the black line, there will not be a large displacement in the lateral position of the rear sensor (relative to the front sensor), so the number of times the car adjusts the servo on the straight line will be significantly reduced, and the stability of the straight line will be better. At the same time, according to the combination of different sensors in the front and rear rows, different cornering strategies are given (reflected in the program in the form of a list) to further improve the stability control ability of the straight line.