Man-vehicle formation driving control method based on markers

Abstract

The invention discloses a marker-based human-vehicle formation driving control method, which belongs to the field of formation driving control, and comprises the following steps: determining a navigator and a follower of a queue, and determining a formation reference object of the follower according to a formation; for the unmanned transport vehicle, identifying the marker of the formation reference object in real time, and obtaining center point offset and man-vehicle distance information based on a visual method; respectively inputting the center point offset and the man-vehicle distance information into a virtual spring model and a fuzzy controller to obtain an expected linear velocity and an expected angular velocity of the unmanned transport vehicle; the expected linear velocity and the expected angular velocity of the unmanned transport vehicle are input into the unmanned transport vehicle kinematic model, and the expected rotating speed of the wheels of the unmanned transport vehicle is obtained; and the wheels are controlled to rotate through the motor controller. The method solves the problems that when an unmanned transport vehicle processor deploys a target recognition model, computing resources are limited, the unmanned transport vehicle is prone to shaking when moving along with a target pedestrian, and the transport efficiency of a single unmanned transport vehicle is low.

Description

Technical Field

[0001] The present invention belongs to the field of platoon driving control, and in particular relates to a vehicle-vehicle platoon driving control method based on markers. Background Art

[0002] At present, the methods of unmanned transport vehicle platooning can be divided into laser radar-based method, active device-based method and vision-based method. However, the laser radar-based method is relatively expensive and has certain difficulties in target identification and locking; the active device-based method requires the user to wear equipment and has a limited detection range. The vision-based method has a low cost and accurate target tracking, and has become the current mainstream solution. However, the current mainstream vision-based method still has some defects.

[0003] Unmanned transport vehicles are autonomous mobile platforms with formation following functions. As a new trend in the development of unmanned transport vehicles, human-vehicle formation driving has the advantages of good passing performance, strong carrying capacity, and reduced labor intensity. At present, the research method for human-vehicle formation driving of unmanned transport vehicles is mainly to use intelligent algorithms to identify the characteristic points of the human body, thereby determining the following target and moving the unmanned transport vehicle towards the target point.

[0004] At present, some scholars have designed a pedestrian following system based on depth camera vision, obtained the distance d between the target pedestrian and the following car and the deviation angle between the virtual spring and the following car, proposed the KCF algorithm and designed the following model motion control strategy.

[0005] The algorithm effectively tracks the target pedestrian in real time and obtains the image position of the target pedestrian. The specific position of the target pedestrian relative to the robot is calculated based on the installation position of the depth camera on the robot and the depth stream information of the depth camera. Then, through the tracking motion control strategy of the virtual spring model, anti-interference and anti-loss pedestrian tracking can be achieved.

[0006] The disadvantages are:

[0007] ① Unmanned transport vehicles use an embedded edge data processing platform with limited computing resources, and the cost of deploying large-scale target recognition models is high;

[0008] ② The angular velocity and linear velocity of the unmanned transport vehicle are coupled together, and the target pedestrian's steps cause random fluctuations when walking, which will cause the unmanned transport vehicle to jitter when following the movement;

[0009] ③The transportation efficiency of a single unmanned transport vehicle is low. Summary of the invention

[0010] In response to the above-mentioned deficiencies in the prior art, the present invention provides a marker-based human-vehicle formation driving control method which solves the problems of limited computing resources when the unmanned transport vehicle processor deploys a target recognition model, the unmanned transport vehicle is prone to jitter when following the target pedestrian, and the low transport efficiency of a single unmanned transport vehicle.

[0011] In order to achieve the above-mentioned purpose of the invention, the technical solution adopted by the present invention is: a method for controlling a vehicle-vehicle platoon based on a marker, comprising:

[0012] Establishing a formation and posting markers on pedestrians and vehicles participating in the formation; determining the leader and follower of the queue, and determining the formation reference object of the follower according to the formation; the marker is a double-ring concentric circle;

[0013] For the unmanned transport vehicle as a follower, the lightweight YOLOv5s target recognition algorithm is used to identify the markers of the formation reference object in real time, and the center point offset M is obtained based on the visual method. o and the distance information l from the formation reference object d ; and offset the center point by M o and the distance information l from the formation reference object d The virtual spring model and the fuzzy controller are input respectively to obtain the expected linear velocity and the expected angular velocity of the unmanned transport vehicle; the expected linear velocity and the expected angular velocity of the unmanned transport vehicle are input into the kinematic model of the unmanned transport vehicle to obtain the expected rotational speed of the wheels of the unmanned transport vehicle; based on the expected rotational speed of the wheels of the unmanned transport vehicle, the rotation of the wheels is controlled by the motor controller.

[0014] Furthermore, the expression of the expected linear speed of the unmanned transport vehicle is:

[0015]

[0016] Among them, v t+1 is the linear velocity at time t+1; v t is the linear velocity at time t; s is the tension state of the virtual spring; F total is the resultant force on the unmanned transport vehicle; m is the mass of the unmanned transport vehicle; d is the differential symbol; t is the time; l d is the distance information to the formation reference object; l s1 is a safe distance; o is the spring relaxation distance; k f is the weighted dynamic damping coefficient; k0 is the elastic coefficient of the virtual spring; α and β are both weight coefficients.

[0017] Furthermore, the resultant force F acting on the unmanned transport vehicle is total The expression is:

[0018]

[0019] F rh =k f |l rh -l s1

[0020]

[0021] F pf =k f |l pf -l s1

[0022]

[0023] Among them, F pri The force received by the priority follower; F fol The combined force received by ordinary followers; F rh is the spring traction force; i is the obstacle number; n is the total number of obstacles; F pd For when l pd ≤l d1 The elastic repulsive force generated when entering the obstacle's influence range; pd The distance between the priority follower and the obstacle when the follower is within the obstacle's influence range; l d1 is the influence range of the obstacle; l rh F is the distance between the priority follower and the formation reference object; pf F is the spring traction force between the ordinary follower and the formation reference object; fd For when l fd ≤l d1 The elastic repulsive force generated when entering the obstacle's influence range; fd is the distance between the ordinary follower and the obstacle when it is within the influence range of the obstacle; j is the number of the adjacent ordinary follower; m is the total number of adjacent ordinary followers; F ff For when l ff ≤l d2 The elastic repulsive force generated when entering the range of the adjacent ordinary follower; l pf is the distance between the ordinary follower and the formation reference object; l ff The distance between a common follower and its adjacent common followers when they are within the influence range; l d2 is the influence range of adjacent ordinary followers; the priority follower is an unmanned transport vehicle whose formation reference object is a pedestrian; and the ordinary follower is an unmanned transport vehicle whose formation reference object is a vehicle.

[0024] Furthermore, the fuzzy controller adopts a two-dimensional fuzzy controller with dual input and single output, and the input of the controller is the center point offset Mo and the distance information l from the formation reference object d , the output is the desired angular velocity w c .

[0025] Furthermore, the center point offset M in the fuzzy controller o The corresponding variable fuzzy set is set as follows: based on the horizontal coordinate pixel value range [0.m m ], including XS fuzzy set [0, m1), VS fuzzy set [m1, m2), S fuzzy set [m2, m3), M fuzzy set [m3, m4), L fuzzy set [m4, m5), VL fuzzy set [m5, m6) and XL fuzzy set [m6, m m ]; m1, m2, m3, m4, m5 and m6 are the center point offsets M o The corresponding variable fuzzy set partitioning threshold;

[0026] The distance information l between the fuzzy controller and the formation reference object d The corresponding variable fuzzy set is set as: based on the range of the distance between the person and the vehicle [0, l m ] are divided into five equally divided fuzzy sets, including VS fuzzy set [0, l1), S fuzzy set [l1, l2), M fuzzy set [l2, l3), L fuzzy set [l3, l4) and VL fuzzy set [l4, l m ]; l1, l2, l3 and l4 are the distance information to the formation reference object. d The corresponding variable fuzzy set partitioning threshold;

[0027] The desired angular velocity w in the fuzzy controller c The corresponding variable fuzzy set is set as follows: based on the angular velocity range [-w m ,w m ] 9 equally divided fuzzy sets, including the NL fuzzy sets whose corresponding values ​​are sorted from small to large [-w m ,w1), NB fuzzy set [w1,w2], NM fuzzy set [w2,w3], NS fuzzy set [w3,0], ZO fuzzy set [0], PS fuzzy set (0,w4], PM fuzzy set (w4,w5], PB fuzzy set (w5,w6] and PL fuzzy set (w6,w m ]; w1, w2, w3, w4, w5 and w6 are all expected angular velocities w c The corresponding variable fuzzy set partitioning threshold.

[0028] Furthermore, the center point offset M oThe corresponding variable fuzzy sets are XS fuzzy set [0, m1), S fuzzy set [m2, m3), M fuzzy set [m3, m4), L fuzzy set [m4, m5) and XL fuzzy set [m6, m m ] and the desired angular velocity w c The corresponding variable fuzzy set NL fuzzy set [-w m ,w1), NS fuzzy set [w3,0), ZO fuzzy set after interval expansion [0-δ, 0+δ], PS fuzzy set (0,w4] and PL fuzzy set (w6,w m ] is a bell-shaped membership function; δ is the expansion difference; the center point offset M o The corresponding variable fuzzy set VS fuzzy set [m1, m2) and VL fuzzy set [m5, m6), and the distance information l of the formation reference object d The corresponding variable fuzzy sets are VS fuzzy set [0, l1), S fuzzy set [l1, l2), M fuzzy set [l2, l3), L fuzzy set [l3, l4) and VL fuzzy set [l4, l m ], and the desired angular velocity w c The membership functions of the corresponding variable fuzzy sets NB fuzzy set [w1, w2], NM fuzzy set [w2, w3], PM fuzzy set (w4, w5], and PB fuzzy set (w5, w6]) are triangular membership functions.

[0029] Furthermore, the fuzzy control rule of the fuzzy controller is:

[0030]

[0031] Among them, H(w c ) is the center point offset M o and the distance information l from the formation reference object d The expected angular velocity w obtained by the corresponding variable fuzzy set c The corresponding variable fuzzy set, the distance information l between the behavior and the formation reference object in the matrix d The corresponding variable fuzzy sets are VL fuzzy set, L fuzzy set, M fuzzy set, S fuzzy set and VS fuzzy set, respectively. The column in the matrix is ​​the center point offset M o The corresponding variable fuzzy sets are XS fuzzy set, VS fuzzy set, S fuzzy set, M fuzzy set, L fuzzy set, VL fuzzy set and XL fuzzy set.

[0032] Furthermore, the fuzzy controller determines the center point offset M based on the Mandani fuzzy inference method. o , distance information l from the formation reference object d and the desired angular velocity w c The ternary fuzzy relation R.

[0033] Furthermore, the center point is offset by M o and the distance information l from the formation reference object d Input the fuzzy controller to obtain the expected angular velocity of the unmanned transport vehicle, which is:

[0034] According to the center point offset M o and the distance information l from the formation reference object d , using the fuzzy controller, the membership value C1 obtained by fuzzy reasoning is obtained; and the membership value C1 is defuzzified using the center of gravity method to obtain the angular velocity of the unmanned transport vehicle;

[0035] It is determined whether the unmanned transport vehicle is in the insensitive area. If so, the expected angular velocity of the unmanned transport vehicle is set to 0; otherwise, the expected angular velocity of the unmanned transport vehicle is set to the angular velocity of the unmanned transport vehicle.

[0036] Furthermore, the expression of the insensitive area is:

[0037]

[0038] Among them, I ss is the insensitive area; i1, i2 and i3 are all constants; l d is the distance information to the formation reference object; e is a natural constant.

[0039] The beneficial effects of the present invention are as follows: the present invention decouples the linear velocity and angular velocity of the unmanned transport vehicle and controls them separately, designs the linear velocity controller by improving the virtual spring model and the weighted dynamic damping coefficient, and designs the angular velocity controller by setting an adaptive "insensitive area" and fuzzy control, thereby effectively solving the problem that the unmanned transport vehicle is prone to shaking, expanding the human-vehicle formation to a human-vehicle-vehicle formation, and can be flexibly configured according to different needs, thereby improving the transportation efficiency of the unmanned transport vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 The figure is a flow chart of the method of the present invention.

[0041] Figure 2 Schematic diagram of a virtual spring model in an embodiment of the present invention.

[0042] Figure 3 Schematic diagram of the relationship between the weighted dynamic damping coefficient and the elastic force in an embodiment of the present invention.

[0043] Figure 4 Schematic diagram of the input and output membership functions of the fuzzy controller in an embodiment of the present invention.

[0044] Figure 5 It is a schematic diagram of the fuzzy controller results in an embodiment of the present invention.

[0045] Figure 6 Schematic diagram of a marker in an embodiment of the present invention.

[0046] Figure 7 Schematic diagram of a virtual spring topology model in an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0048] like Figure 1 As shown, in one embodiment of the present invention, a method for controlling a vehicle-vehicle platoon based on a marker includes:

[0049] Establishing a formation and posting markers on pedestrians and vehicles participating in the formation; determining the leader and follower of the queue, and determining the formation reference object of the follower according to the formation; the marker is a double-ring concentric circle;

[0050] For the unmanned transport vehicle as a follower, the lightweight YOLOv5s target recognition algorithm is used to identify the markers of the formation reference object in real time, and the center point offset M is obtained based on the visual method. o and the distance information l from the formation reference object d ; and offset the center point by M o and the distance information l from the formation reference object d The virtual spring model and the fuzzy controller are input respectively to obtain the expected linear velocity and the expected angular velocity of the unmanned transport vehicle; the expected linear velocity and the expected angular velocity of the unmanned transport vehicle are input into the kinematic model of the unmanned transport vehicle to obtain the expected rotational speed of the wheels of the unmanned transport vehicle; based on the expected rotational speed of the wheels of the unmanned transport vehicle, the rotation of the wheels is controlled by the motor controller.

[0051] In this embodiment, the network structure that the lightweight YOLOv5s target recognition algorithm relies on is specifically: the backbone network of the YOLOv5s network is replaced by the MobileNetV3 network. Figure 6 The figure shows a schematic diagram of double concentric circles.

[0052] like Figure 2 As shown, it is a mechanical diagram of the virtual spring model of the priority follower under compression, relaxation and tension states.

[0053] The expression of the expected linear speed of the unmanned transport vehicle is:

[0054]

[0055] Among them, v t+1 is the linear velocity at time t+1; v t is the linear velocity at time t; s is the tension state of the virtual spring; F total is the resultant force on the unmanned transport vehicle; m is the mass of the unmanned transport vehicle; d is the differential symbol; t is the time; l d is the distance information to the formation reference object; l s1 is a safe distance; o is the spring relaxation distance; k f is the weighted dynamic damping coefficient; k0 is the elastic coefficient of the virtual spring; α and β are both weight coefficients.

[0056] In this embodiment, Figure 3 This is a diagram illustrating the relationship between distance and elastic force when adjusting the weight in the calculation formula of the weighted dynamic damping coefficient.

[0057] The resultant force F on the unmanned transport vehicle total The expression is:

[0058]

[0059] F rh =k f |l rh -l s1

[0060]

[0061]

[0062] F pf =k f |l pf -l s1

[0063]

[0064] Among them, F pri The force received by the priority follower; F fol The combined force received by ordinary followers; F rh is the spring traction force; i is the obstacle number; n is the total number of obstacles; F pd For when l pd ≤l d1 The elastic repulsive force generated when entering the obstacle's influence range; pd The distance between the priority follower and the obstacle when the follower is within the obstacle's influence range; l d1is the influence range of the obstacle; l rh F is the distance between the priority follower and the formation reference object; pf F is the spring traction force between the ordinary follower and the formation reference object; fd For when l fd ≤l d1 The elastic repulsive force generated when entering the obstacle's influence range; fd is the distance between the ordinary follower and the obstacle when it is within the influence range of the obstacle; j is the number of the adjacent ordinary follower; m is the total number of adjacent ordinary followers; F ff For when l ff ≤l d2 The elastic repulsive force generated when entering the range of the adjacent ordinary follower; l pf is the distance between the ordinary follower and the formation reference object; l ff The distance between a common follower and its adjacent common followers when they are within the influence range; l d2 is the influence range of adjacent ordinary followers; the priority follower is an unmanned transport vehicle whose formation reference object is a pedestrian; and the ordinary follower is an unmanned transport vehicle whose formation reference object is a vehicle.

[0065] like Figure 7 As shown, this is a schematic diagram of the virtual spring topology of human-vehicle-vehicle.

[0066] The fuzzy controller adopts a two-dimensional fuzzy controller with dual input and single output. The input of the controller is the center point offset M o and the distance information l from the formation reference object d , the output is the desired angular velocity w c .

[0067] The center point offset M in the fuzzy controller o The corresponding variable fuzzy set is set as follows: based on the horizontal coordinate pixel value range [0.m m ], including XS fuzzy set [0, m1), VS fuzzy set [m1, m2), S fuzzy set [m2, m3), M fuzzy set [m3, m4), L fuzzy set [m4, m5), VL fuzzy set [m5, m6) and XL fuzzy set [m6, m m ]; m1, m2, m3, m4, m5 and m6 are the center point offsets M o The corresponding variable fuzzy set partitioning threshold;

[0068] The distance information l between the fuzzy controller and the formation reference object d The corresponding variable fuzzy set is set as: based on the range of the distance between the person and the vehicle [0, l m] are divided into five equally divided fuzzy sets, including VS fuzzy set [0, l1), S fuzzy set [l1, l2), M fuzzy set [l2, l3), L fuzzy set [l3, l4) and VL fuzzy set [l4, l m ]; l1, l2, l3 and l4 are the distance information to the formation reference object. d The corresponding variable fuzzy set partitioning threshold;

[0069] The desired angular velocity w in the fuzzy controller c The corresponding variable fuzzy set is set as follows: based on the angular velocity range [-w m ,w m ] 9 equally divided fuzzy sets, including the NL fuzzy sets whose corresponding values ​​are sorted from small to large [-w m ,w1), NB fuzzy set [w1,w2], NM fuzzy set [w2,w3], NS fuzzy set [w3,0], ZO fuzzy set [0], PS fuzzy set (0,w4], PM fuzzy set (w4,w5], PB fuzzy set (w5,w6] and PL fuzzy set (w6,w m ]; w1, w2, w3, w4, w5 and w6 are all expected angular velocities w c The corresponding variable fuzzy set partitioning threshold.

[0070] The center point offset M o The corresponding variable fuzzy sets are XS fuzzy set [0, m1), S fuzzy set [m2, m3), M fuzzy set [m3, m4), L fuzzy set [m4, m5) and XL fuzzy set [m6, m m ] and the desired angular velocity w c The corresponding variable fuzzy set NL fuzzy set [-w m ,w1), NS fuzzy set [w3,0), ZO fuzzy set after interval expansion [0-δ, 0+δ], PS fuzzy set (0,w4] and PL fuzzy set (w6,w m ] is a bell-shaped membership function; δ is the expansion difference; the center point offset M o The corresponding variable fuzzy set VS fuzzy set [m1, m2) and VL fuzzy set [m5, m6), and the distance information l of the formation reference object d The corresponding variable fuzzy sets are VS fuzzy set [0, l1), S fuzzy set [l1, l2), M fuzzy set [l2, l3), L fuzzy set [l3, l4) and VL fuzzy set [l4, l m ], and the desired angular velocity w c The membership functions of the corresponding variable fuzzy sets NB fuzzy set [w1, w2], NM fuzzy set [w2, w3], PM fuzzy set (w4, w5], and PB fuzzy set (w5, w6]) are triangular membership functions.

[0071] The fuzzy control rule of the fuzzy controller is:

[0072]

[0073] Among them, H(w c ) is the center point offset M o and the distance information l from the formation reference object d The expected angular velocity w obtained by the corresponding variable fuzzy set c The corresponding variable fuzzy set, the distance information l between the behavior and the formation reference object in the matrix d The corresponding variable fuzzy sets are VL fuzzy set, L fuzzy set, M fuzzy set, S fuzzy set and VS fuzzy set, respectively. The column in the matrix is ​​the center point offset M o The corresponding variable fuzzy sets are XS fuzzy set, VS fuzzy set, S fuzzy set, M fuzzy set, L fuzzy set, VL fuzzy set and XL fuzzy set.

[0074] The fuzzy controller determines the center point offset M based on the Mandani fuzzy inference method. o , distance information l from the formation reference object d and the desired angular velocity w c The ternary fuzzy relation R.

[0075] In this embodiment, taking the human-vehicle formation (pedestrian-priority follower) as an example, the fuzzy controller construction process is explained: using the marker center point offset M o Distance information between people and vehicles rh To control the angular velocity of the carrier platform; the center point offset M o Refers to the absolute value of the difference between the horizontal coordinate pixel point (X-pixel) of the marker in the depth camera and the center point. For example, the range of the D455 depth camera used in the present invention is [0, 1280] dp, 640 dp is the center point, when the marker is in the range of [0, 640] dp, the angular velocity is negative, and the carrier platform turns left; when the marker is in the range of [640, 1280] dp, the angular velocity is positive, and the carrier platform turns right. When the distance between the person and the vehicle is constant, the offset is positively correlated with the angular velocity; when the offset is constant, the distance between the person and the vehicle is also positively correlated with the angular velocity.

[0076] A two-dimensional fuzzy controller with dual input and single output is used. The input of the controller is the center point offset M. o Distance information between people and vehicles rh , the output is the angular velocity w c The specific design steps of the fuzzy controller are as follows:

[0077] ① Define the fuzzy set of input and output variables. The horizontal coordinate pixel value range is [0, 1280] dp. Set it to 7 fuzzy sets, namely XS (extremely small), VS (very small), S (small), M (medium), L (large), VL (very large), XL (extremely large); the distance between the person and the vehicle is l rh The value range is [0, 10]m, and it is set to 5 fuzzy sets, namely VS (very small), S (small), M (medium), L (large), VL (very large); the angular velocity w c The value range is [-0.3, 0.3] rad / s, and it is set to 9 fuzzy sets, namely NL (negative maximum), NB (negative large), NM (negative medium), NS (negative small), ZO (zero), PS (positive small), PM (positive medium), PB (positive large), and PL (positive maximum).

[0078] ②Define the input and output membership functions and the center point offset M o The membership functions of XS (extremely small), S (small), M (medium), L (large), and XL (maximum) are set to gbellmf (bell-shaped membership function), and the angular velocity w c The membership functions of NL (negative maximum), NS (negative small), ZO (zero), PS (positive small), and PL (positive maximum) are set to gbellmf (bell-shaped membership function), and the interval of ZO (zero) is expanded. The membership functions of the remaining fuzzy sets are set to trimf (triangular membership function), such as Figure 4 and Figure 5 shown.

[0079] ③Establish fuzzy control rules:

[0080] If(M o is XS)and(l rh is VL)then(w c is NL);

[0081] If the center point offset M o is XS (extremely small) and the distance between the person and the vehicle is l rh is VL (very large) then the angular velocity w c Just take NL (extremely negative);

[0082] If(M o is VS)and(l rh is VL)then(w c is NB);

[0083] If the center point offset M o is VS (very small) and the distance between the person and the vehicle is l rh is VL (very large) then the angular velocity w c Just take NB (negative big);

[0084] If(M o is S)and(l rh is VL)then(w c is NS);

[0085] If the center point offset M o is S (small) and the distance between the person and the vehicle is l rh is VL (very large) then the angular velocity w c Just take NS (negative small);

[0086] …

[0087] ④ Establish a fuzzy control table (as shown in Table 1)

[0088] Table 1

[0089]

[0090] ⑤ Fuzzy reasoning. In the designed fuzzy controller, the relationship of the fuzzy reasoning statement "If A and B then C" is (A∧B→C). Using the Mandani fuzzy reasoning method, the ternary fuzzy relationship R is determined as:

[0091] R=(A×B) T1 ×C

[0092] In the formula, (A×B) T1 It is composed of A×B, with a size of m×n, where n and m are the number of elements in the domain of A and B respectively; A is the center point offset M o The domain,set is [XS,VS,S,M,L,VL,XL], and the membership value of a certain center point offset is determined to be [a1,a2,a3,a4,a5,a6,a7,] after multiple adjustments based on the previous membership function; B is the distance between the person and the vehicle l rh The domain, the set is [VL, L, M, S, VS], and the membership value of a certain distance between a person and a car is determined to be [b1, b2, b3, b4, b5] after multiple adjustments based on the previous membership function; C is the angular velocity w c The domain,set is [NL, NB, NM, NS, ZO, PS, PM, PB, PL]. The membership value corresponding to the center point offset and the distance between the person and the vehicle is determined to be [c1,c2,c3,c4,c5,c6,c7,c8,c9] after multiple adjustments of the previous membership function.

[0093] The formula for calculating (A×B) is as follows:

[0094]

[0095] In the formula, m(a1,b1) means taking the smaller value of a1 and b1.

[0096] Expand (A×B) to (A×B) T1 The column vector looks like this:

[0097] (A×B) T1 =[m(a1,b1)m(a1,b2)m(a1,b3)m(a1,b4)m(a1,b5)m(a2,b1)…m(a7,b5)] T

[0098] Therefore, the calculation formula of the ternary fuzzy relationship R is as follows:

[0099]

[0100] In the formula, m(a1,b1,c1) means taking the smaller value among a1,b1,c1.

[0101] According to the fuzzy inference rules and R, the C1 corresponding to the given input A1 and B1 can be obtained as:

[0102] C1=(A1×B1) T2 ×R

[0103] Where A1 is the membership value corresponding to the actual measured center point offset [d1, d2, d3, d4, d5, d6, d7]; B1 is the membership value corresponding to the actual measured distance between the person and the vehicle [e1, e2, e3, e4, e5]; C1 is the membership value obtained by fuzzy reasoning after the actual measured center point offset and the distance between the person and the vehicle [f1, f2, f3, f4, f5, f6, f7, f8, f9]; T2 is the row-column vector conversion.

[0104] The formula (A1×B1) is as follows:

[0105]

[0106] In the formula, m(d1,e1) means taking the smaller value of d1 and e1.

[0107] Expand (A1×B1) to (A1×B1) T2 The row vector looks like this:

[0108] (A1×B1) T2 =[m(d1,e1)m(d1,e2)m(d1,e3)m(d1,e4)m(d1,e5)m(d2,e1)…m(d7,e5)]

[0109] According to the actual input A1 and B1, the inference C1 is the membership value [f1,f2,f3,f4,f5,f6,f7,f8,f9]. It is also necessary to perform defuzzification to obtain the angular velocity w of the carrier platform.c The exact value of . The centroid method is often used for defuzzification, and the formula is as follows:

[0110]

[0111] Where μ(w) is the output membership function; w is the specific value corresponding to the output membership function.

[0112] This gives the center point offset M o Distance information between people and vehicles rh The angular velocity of the carrier platform w c .

[0113] Offset the center point by M o and the distance information l from the formation reference object d Input the fuzzy controller to obtain the expected angular velocity of the unmanned transport vehicle, which is:

[0114] According to the center point offset M o and the distance information l from the formation reference object d , using the fuzzy controller, the membership value C1 obtained by fuzzy reasoning is obtained; and the membership value C1 is defuzzified using the center of gravity method to obtain the angular velocity of the unmanned transport vehicle;

[0115] It is determined whether the unmanned transport vehicle is in the insensitive area. If so, the expected angular velocity of the unmanned transport vehicle is set to 0; otherwise, the expected angular velocity of the unmanned transport vehicle is set to the angular velocity of the unmanned transport vehicle.

[0116] The expression of the insensitive area is:

[0117]

[0118] Among them, I ss is the insensitive area; i1, i2 and i3 are all constants; l d is the distance information to the formation reference object; e is a natural constant.

Claims

1. A method for controlling a vehicle-vehicle platoon based on a marker, characterized in that: include: Establish a formation and post signs on pedestrians and vehicles participating in the formation; Determine the leader and follower of the queue, and determine the formation reference object of the follower according to the formation; the marker is a double-ring concentric circle; For the unmanned transport vehicle as a follower, the lightweight YOLOv5s target recognition algorithm is used to identify the markers of the formation reference object in real time, and the center point offset M is obtained based on the visual method. o and the distance information l from the formation reference object d ; and offset the center point by M o and the distance information l from the formation reference object d The virtual spring model and the fuzzy controller are input respectively to obtain the expected linear velocity and the expected angular velocity of the unmanned transport vehicle; the expected linear velocity and the expected angular velocity of the unmanned transport vehicle are input into the kinematic model of the unmanned transport vehicle to obtain the expected rotational speed of the wheels of the unmanned transport vehicle; based on the expected rotational speed of the wheels of the unmanned transport vehicle, the rotation of the wheels is controlled by the motor controller.

2. The method for controlling a vehicle-vehicle platoon based on markers according to claim 1, characterized in that: The expression of the expected linear speed of the unmanned transport vehicle is: Among them, v t+1 is the linear velocity at time t+1; v t is the linear velocity at time t; s is the tension state of the virtual spring; F total is the resultant force on the unmanned transport vehicle; m is the mass of the unmanned transport vehicle; d is the differential symbol; t is the time; l d is the distance information to the formation reference object; l s1 is a safe distance; o is the spring relaxation distance; k f is the weighted dynamic damping coefficient; k0 is the elastic coefficient of the virtual spring; α and β are both weight coefficients.

3. The method for controlling a vehicle-vehicle platoon based on markers according to claim 2, characterized in that: The resultant force F on the unmanned transport vehicle total The expression is: F rh =k f |l rh -l s1 F pf =k f |l pf -l s1 Among them, F pri The force received by the priority follower; F fol The combined force received by ordinary followers; F rh is the spring traction force; i is the obstacle number; n is the total number of obstacles; F pd For when l pd ≤l d1 The elastic repulsive force generated when entering the obstacle's influence range; pd The distance between the priority follower and the obstacle when the follower is within the obstacle's influence range; l d1 is the influence range of the obstacle; l rh F is the distance between the priority follower and the formation reference object; pf F is the spring traction force between the ordinary follower and the formation reference object; fd For when l fd ≤l d1 The elastic repulsive force generated when entering the obstacle's influence range; fd is the distance between the ordinary follower and the obstacle when it is within the influence range of the obstacle; j is the number of the adjacent ordinary follower; m is the total number of adjacent ordinary followers; F ff For when l ff ≤l d2 The elastic repulsive force generated when entering the range of the adjacent ordinary follower; l pf is the distance between the ordinary follower and the formation reference object; l ff The distance between a common follower and its adjacent common followers when they are within the influence range; l d2 is the influence range of adjacent ordinary followers; the priority follower is an unmanned transport vehicle whose formation reference object is a pedestrian; and the ordinary follower is an unmanned transport vehicle whose formation reference object is a vehicle.

4. The method for controlling a vehicle-vehicle platoon based on markers according to claim 1, characterized in that: The fuzzy controller adopts a two-dimensional fuzzy controller with dual input and single output. The input of the controller is the center point offset M o and the distance information l from the formation reference object d , the output is the desired angular velocity w c .

5. The method for controlling the driving of a vehicle-vehicle platoon based on markers according to claim 4 is characterized in that: The center point offset M in the fuzzy controller o The corresponding variable fuzzy set is set as follows: based on the horizontal coordinate pixel value range [0.m m ], including XS fuzzy set [0, m1), VS fuzzy set [m1, m2), S fuzzy set [m2, m3), M fuzzy set [m3, m4), L fuzzy set [m4, m5), VL fuzzy set [m5, m6) and XL fuzzy set [m6, m m ]; m1, m2, m3, m4, m5 and m6 are the center point offsets M o The corresponding variable fuzzy set partitioning threshold; The distance information l between the fuzzy controller and the formation reference object d The corresponding variable fuzzy set is set as: based on the range of the distance between the person and the vehicle [0, l m ] are divided into five equally divided fuzzy sets, including VS fuzzy set [0, l1), S fuzzy set [l1, l2), M fuzzy set [l2, l3), L fuzzy set [l3, l4) and VL fuzzy set [l4, l m ]; l1, l2, l3 and l4 are the distance information to the formation reference object. d The corresponding variable fuzzy set partitioning threshold; The desired angular velocity w in the fuzzy controller c The corresponding variable fuzzy set is set as follows: based on the angular velocity range [-w m ,w m ] 9 equally divided fuzzy sets, including the NL fuzzy sets whose corresponding values ​​are sorted from small to large [-w m ,w1), NB fuzzy set [w1,w2], NM fuzzy set [w2,w3], NS fuzzy set [w3,0], ZO fuzzy set [0], PS fuzzy set (0,w4], PM fuzzy set (w4,w5], PB fuzzy set (w5,w6] and PL fuzzy set (w6,w m ]; w1, w2, w3, w4, w5 and w6 are all expected angular velocities w c The corresponding variable fuzzy set partitioning threshold.

6. The method for controlling the driving of a vehicle-vehicle platoon based on markers according to claim 5, characterized in that: The center point offset M o The corresponding variable fuzzy sets are XS fuzzy set [0, m1), S fuzzy set [m2, m3), M fuzzy set [m3, m4), L fuzzy set [m4, m5) and XL fuzzy set [m6, m m ] and the desired angular velocity w c The corresponding variable fuzzy set NL fuzzy set [-w m ,w1), NS fuzzy set [w3,0), ZO fuzzy set after interval expansion [0-δ, 0+δ], PS fuzzy set (0,w4] and PL fuzzy set (w6,w m ] is a bell-shaped membership function; δ is the expansion difference; the center point offset M o The corresponding variable fuzzy set VS fuzzy set [m1, m2) and VL fuzzy set [m5, m6), and the distance information l of the formation reference object d The corresponding variable fuzzy sets are VS fuzzy set [0, l1), S fuzzy set [l1, l2), M fuzzy set [l2, l3), L fuzzy set [l3, l4) and VL fuzzy set [l4, l m ], and the desired angular velocity w c The membership functions of the corresponding variable fuzzy sets NB fuzzy set [w1, w2], NM fuzzy set [w2, w3], PM fuzzy set (w4, w5], and PB fuzzy set (w5, w6]) are triangular membership functions.

7. The method for controlling a vehicle-vehicle platoon based on markers according to claim 5, characterized in that: The fuzzy control rule of the fuzzy controller is: Among them, H(w c ) is the center point offset M o and the distance information l from the formation reference object d The expected angular velocity w obtained by the corresponding variable fuzzy set c The corresponding variable fuzzy set, the distance information l between the behavior and the formation reference object in the matrix d The corresponding variable fuzzy sets are VL fuzzy set, L fuzzy set, M fuzzy set, S fuzzy set and VS fuzzy set, respectively. The column in the matrix is ​​the center point offset M o The corresponding variable fuzzy sets are XS fuzzy set, VS fuzzy set, S fuzzy set, M fuzzy set, L fuzzy set, VL fuzzy set and XL fuzzy set.

8. The method for controlling a vehicle-vehicle platoon based on markers according to claim 4, characterized in that: The fuzzy controller determines the center point offset M based on the Mandani fuzzy inference method. o , distance information l from the formation reference object d and the desired angular velocity w c The ternary fuzzy relation R.

9. The method for controlling a vehicle-vehicle platoon based on markers according to claim 1, characterized in that: Offset the center point by M o and the distance information l from the formation reference object d Input the fuzzy controller to obtain the expected angular velocity of the unmanned transport vehicle, which is: According to the center point offset M o and the distance information l from the formation reference object d , using the fuzzy controller, the membership value C1 obtained by fuzzy reasoning is obtained; and the membership value C1 is defuzzified using the center of gravity method to obtain the angular velocity of the unmanned transport vehicle; It is determined whether the unmanned transport vehicle is in the insensitive area. If so, the expected angular velocity of the unmanned transport vehicle is set to 0; otherwise, the expected angular velocity of the unmanned transport vehicle is set to the angular velocity of the unmanned transport vehicle.

10. The method for controlling the driving of a vehicle-vehicle platoon based on markers according to claim 9, characterized in that: The expression of the insensitive area is: Among them, I ss is the insensitive area; i1, i2 and i3 are all constants; l d is the distance information to the formation reference object; e is a natural constant.

Images