Design method of multi-axis unmanned vehicle chassis based on ROS general robot operating system

Abstract

The application discloses a design method of a multi-axle unmanned vehicle chassis based on a ROS general robot operating system, and comprises the following steps: step one, designing a multi-axle vehicle electric drive hardware platform, S1.1 hardware architecture design; S1.2 software platform design; step two, designing a multi-axle vehicle steering and torque distribution algorithm based on target speed control, S2.1 calculating multi-axle vehicle wheel steering angles by using an Ackerman steering angle principle; S2.2 designing a fuzzy torque control method according to different steering inside and outside speeds; and step three, designing a control algorithm module for controlling vehicle steering. The torque distribution algorithm based on target speed control can effectively distribute the multi-wheel torque of the system, guarantee the basic stability of the target speed, realize the fuzzy self-distribution strategy of the torque, has good practical value, can be used for realizing the algorithm basis of unmanned driving, and has the characteristics of good torque distribution effect of target speed control and strong practicability.

Description

Technical Field

[0001] This invention relates to the field of automotive technology, and more specifically to a design method for a multi-axis unmanned vehicle chassis based on the ROS general robot operating system. Background Technology

[0002] With the deepening development of human civilization and industrialization, heavy vehicles and special vehicles are becoming more and more common. The steering and torque distribution strategies of multi-axle vehicles have become a major factor affecting vehicles. In particular, in recent years, the rapid development of electric vehicles and the gradual application of unmanned driving robotic systems have put forward new requirements for the distribution strategies of multi-axle special vehicles under various driving conditions.

[0003] Meanwhile, due to the large number of cars and the latency of control, car accidents are arguably the most common type of transportation. The world has placed higher demands on the safety performance of cars and on their control algorithms. Electric vehicles provide the technological possibility for meeting these demands. The inherent distributed system architecture and the superior efficiency of electric drive are far superior to those of internal combustion engine cars. Future autonomous driving will also be based on this low-latency, high-efficiency motor architecture, thus making the research on electric vehicles necessary.

[0004] In the existing technology, there are no relevant records or effective solutions for the steering and torque distribution problems of the aforementioned unmanned multi-axle vehicles. Therefore, it is urgent to design a design method for a multi-axle unmanned vehicle chassis based on the ROS general robot operating system to solve the steering and torque distribution problems of the aforementioned unmanned multi-axle vehicles. Summary of the Invention

[0005] To address the aforementioned problems, this invention aims to provide a design method for a multi-axis unmanned vehicle chassis based on the ROS general robot operating system. This method, through the design of a torque distribution algorithm based on target speed control, can effectively distribute the torque of the system across multiple wheels, ensuring the basic stability of the target speed. It realizes a fuzzy self-allocation strategy for torque, has good practical value, and can be used as the algorithmic basis for realizing unmanned driving. It features good torque distribution effect based on target speed control and strong practicality.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] Design methodology for multi-axis unmanned vehicle chassis based on ROS general robot operating system, including

[0008] Step 1: Design the hardware platform for multi-axle vehicle electric drive

[0009] S1.1 Hardware Architecture Design;

[0010] S1.2 software platform design;

[0011] Step 2: Design a multi-axle vehicle steering and torque distribution algorithm based on target speed control.

[0012] S2.1 Calculate the wheel steering angle of a multi-axle vehicle using the Ackermann steering principle;

[0013] S2.2 Design a fuzzy torque control method based on the different speeds inside and outside the steering wheel;

[0014] Step 3: Design the control algorithm module for controlling vehicle steering.

[0015] Preferably, the hardware architecture design includes

[0016] (1) A three-level embedded hardware platform is used, and different processors are used in each level. The Nano processor is used as the master controller of the multi-axis homing chassis, and the PC is used as the slave.

[0017] (2) The whole system combines multiple bus structures such as local area network, USB interface, and CAN bus. The PC and the underlying Jetson Nano processor use the Dandelion remote networking software to build a local area network under the Internet and exchange data and remote control through the TCP / IP protocol.

[0018] Preferably, the software platform design process includes:

[0019] The lowest-level ROS chassis control software is a software system used to directly control vehicle chassis information and acquire chassis information;

[0020] The middle layer is a remote local area network built on the Dandelion VPN software, which is the information interaction layer of the entire system;

[0021] The top layer is the decision-making layer, which is the host computer layer where humans are in the loop or the machine controls itself. It is the central hub for issuing instructions to the entire system.

[0022] Preferably, the process of calculating the wheel steering angle of a multi-axle vehicle using the Ackermann steering principle includes:

[0023] S2.11 Calculate the wheel steering angle

[0024] Let K be the center distance between the steering wheels and L be the wheelbase of the electric vehicle. The formula for the wheel steering angle, derived geometrically, is as follows:

[0025]

[0026] Where α1 is the steering angle of the left front wheel and α2 is the steering angle of the right front wheel;

[0027] S2.12 Calculates the relationship between the wheel steering angle and the Ackermann steering angle based on the calculated steering angle of the multi-axle vehicle.

[0028] Preferably, the calculation process for the relationship between the wheel angle and the Ackermann angle in step S2.12 includes:

[0029] (1) Let angle a0 be a known angle. The hypotenuse of the triangle is perpendicular to the direction of the wheel. We can then conclude that:

[0030] a = 90 - a0 (2)

[0031] f0 = 90 - f (3)

[0032] (2) Figure 7 As shown, given that the rotation angle of axle one is 'a', the distance of axle one from the midpoint is 'A', and the distance of axle two from the midpoint is 'B', let the distance of the midpoint from the turning center be 'S', and the rotation angle of axle two be 'b'. Given the distance of axle two from the midpoint is 'B', we can deduce that:

[0033]

[0034]

[0035]

[0036]

[0037]

[0038]

[0039]

[0040] (3) Given that a0 is the program turning angle, L is the wheel track, and A+B is the distance from the axle to the turning point, we can deduce:

[0041]

[0042]

[0043]

[0044] Therefore, given the rotation angle of one axle and the distances between the midpoints of the turns of each axle, the rotation angles of multiple axles can be calculated.

[0045] Preferably, the process of designing the fuzzy torque control method based on the difference between the inside and outside speeds during steering includes:

[0046] S2.21 Design torque distribution control algorithm;

[0047] S2.22 processes the feedback speed to achieve differential control of the left and right wheels.

[0048] Preferably, the design process of the torque distribution control algorithm includes:

[0049] The longitudinal vehicle motion is based on a PD-controlled overall vehicle speed control and torque distribution strategy, leveraging the advantage that multiple wheels are individually driven. A human-in-the-loop control is employed to manage the target speed, and the algorithm calculates the torque distribution value to control the torque output of multiple wheels throughout the vehicle. The torque distribution formula is as follows:

[0050] T=K P ×ΔV+K d ×(ΔV-V P )+T0 (14)

[0051]

[0052]

[0053] Because the car's steering is in an arc, the turning speed ratio is:

[0054]

[0055]

[0056]

[0057]

[0058]

[0059] By analogy, the steering angles of the multiple wheels turning outwards can be derived as follows:

[0060]

[0061]

[0062] Preferably, the process of processing the feedback speed includes:

[0063] Let the feedback speed be V0, the left wheel speed be V2, the right wheel speed be V1, and the car be turning right with the left wheel being the outermost wheel. Then we have:

[0064]

[0065]

[0066]

[0067]

[0068]

[0069]

[0070] a×V0=V1 (30)

[0071]

[0072] b×V0=V2 (32)

[0073]

[0074] Therefore, the algorithm for processing the real-time speed of the right wheel is V0*2 / (1+R2 / R1);

[0075] The algorithm for processing the real-time speed of the revolver is V0*2 / (1+R1 / R2);

[0076] Furthermore, when turning from right to left, the right wheel becomes the outer wheel, and the processing algorithm can handle the interaction between the left and right sides.

[0077] Preferably, the design process of the control algorithm module includes:

[0078] S3.1 Design S calculation module;

[0079] S3.2 Design Control System

[0080] S3.21 Design target speed receiving module;

[0081] S3.22 designs a torque PD controller.

[0082] Preferably, the design process of the torque PD controller includes:

[0083] (1) Design the left front wheel control algorithm module;

[0084] (2) Design the left rear wheel control algorithm module;

[0085] (3) Design module torque distribution module.

[0086] The beneficial effects of this invention are: This invention discloses a design method for a multi-axis unmanned vehicle chassis based on the ROS general robot operating system. Compared with the prior art, the improvement of this invention lies in:

[0087] (1) This invention designs a multi-axis unmanned vehicle chassis based on the ROS general robot operating system. This method can effectively distribute the torque of the system's multiple wheels by designing a torque distribution algorithm based on target speed control, ensuring the basic stability of the target speed and realizing the fuzzy self-allocation strategy of torque. It has good practical value and can be used as the algorithm basis for realizing unmanned driving.

[0088] (2) The method designs a multi-axis chassis based on the ROS robot system, which can serve as a complete experimental platform to verify the motion algorithm of multi-axis special vehicles. It can better reflect the motion characteristics of multi-axis vehicles and has certain significance for experimental applications and teaching. More in-depth theoretical exploration and engineering practice can be carried out based on this platform. At the same time, it can be used for the later unmanned test vehicle. Combined with the chassis control program, it can realize multiple functions such as parallel driving. It has the advantages of good torque distribution effect of target speed control and strong practicality. Attached Figure Description

[0089] Figure 1 This is a flowchart illustrating the design process of the multi-axis unmanned vehicle chassis design method based on the ROS general robot operating system of this invention.

[0090] Figure 2 This is a diagram of the complete vehicle hardware architecture of the present invention.

[0091] Figure 3 This is a network topology diagram of the hardware platform of the present invention.

[0092] Figure 4 This is a software architecture diagram of the present invention.

[0093] Figure 5 This is a schematic diagram of the Ackerman turn principle of the present invention.

[0094] Figure 6 This is a diagram showing the relationship between the wheel and the Ackermann steering angle of the present invention.

[0095] Figure 7 This is an Ackermann cornering diagram of the six wheels on the right side of this invention.

[0096] Figure 8 This is a diagram showing the turning angles of the wheels on both sides of the present invention.

[0097] Figure 9 This is a diagram showing the torque distribution R-side relationship of the present invention.

[0098] Figure 10 This is the overall Simulink program flowchart of the present invention.

[0099] Figure 11 This is a diagram of the S-calculation module of the present invention.

[0100] Figure 12This is an internal block diagram of the vehicle control module of the present invention.

[0101] Figure 13 This is a diagram of the vehicle chassis information receiving module of the present invention.

[0102] Figure 14 This is a diagram of the control algorithm module for the left front wheel of a vehicle according to the present invention.

[0103] Figure 15 This is a block diagram of the Ackerman right turn algorithm of the present invention.

[0104] Figure 16 This is a diagram of the control algorithm module for the left rear wheel of a vehicle according to the present invention.

[0105] Figure 17 This is a schematic diagram of the algorithm for distributing the left rear wheel of a vehicle according to the present invention.

[0106] Figure 18 This is a schematic diagram of the target speed processing algorithm for the non-starting wheel of the present invention.

[0107] Figure 19 This is a schematic diagram of the torque distribution module of a vehicle according to the present invention.

[0108] Figure 20 This is a response effect diagram for Embodiment 2 of the present invention with a real-time speed input of 5 revolutions per minute.

[0109] Figure 21 The diagram shows the response effect of a real-time speed input of 5 revolutions per minute and a turning angle of 60 degrees in Embodiment 2 of the present invention.

[0110] Figure 22 This is a diagram showing the response effect of the P controller in a real vehicle frame experiment according to Embodiment 2 of the present invention.

[0111] Figure 23 This is a response effect diagram of PD control at a real-time speed of 30 revolutions per minute in Embodiment 2 of the present invention.

[0112] Figure 24 This is a response effect diagram of the P controller in an outdoor vehicle experiment according to Embodiment 2 of the present invention.

[0113] Figure 25 This is a vehicle trajectory diagram from Embodiment 2 of the present invention.

[0114] Figure 26 This is a response effect diagram of the PD controller in an outdoor test vehicle according to Embodiment 2 of the present invention. Detailed Implementation

[0115] To enable those skilled in the art to better understand the technical solutions of the present invention, the technical solutions of the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0116] Example 1: Refer to Appendix Figure 1-26 The design method for a multi-axis unmanned vehicle chassis based on the ROS general robot operating system is shown in this embodiment. This example uses a six-axis unmanned vehicle chassis, including...

[0117] Step 1: Design the hardware platform for multi-axle vehicle electric drive

[0118] S1.1 Hardware Architecture Design

[0119] (1) As Figure 2 As shown, the hardware platform designed in this embodiment is a three-tiered embedded design scheme, with each layer using different processors. It employs a PC and a Nano as the master and slave devices, forming an open hardware control scheme that leaves room for hardware expansion in scientific research.

[0120] The Nano processor, acting as the main controller of the six-axis CAN chassis, parses the chassis CAN messages acquired by the CAN analyzer and publishes them to the nodes of the ROS system.

[0121] The PC acts as a slave device, obtaining ROS node information through a virtual local area network to perform ROS system operations such as remote control and autonomous positioning and navigation.

[0122] (2) Figure 3 As shown, the entire system combines multiple bus structures such as local area network, USB interface, and CAN bus. The PC and the underlying Jetson Nano processor use the Pomegranate remote networking software to build a local area network under the Internet, and exchange data and remote control through the TCP / IP protocol.

[0123] The Jetson Nano host connects and exchanges data with the CAN network via a USB interface, and its internal ROS system communicates using the TCP / IP protocol based on distributed control.

[0124] The CAN network is formed by combining the six CAN buses of three CAN bus analyzers to form a CAN bus network and a communication loop. One channel receives status information and the other controls the vehicle's movement. Among them, the wheel torque, vehicle speed, steering mechanism, power battery information, etc. are CAN network nodes used for feedback and control.

[0125] The GPS / inertial combination device connects to the Jetson Nano host via a CAN-to-USB interface to exchange data and monitor the operating data of the six-axle vehicle;

[0126] S1.2 Software Platform Design

[0127] like Figure 4 As shown, the software platform designed in this embodiment is a hierarchical control system:

[0128] The lowest-level ROS chassis control software is a software system used to directly control and acquire vehicle chassis information. It is the core and foundation of the entire system. It mainly uses ROS system's Msg and topic services to build and provide them to other nodes. The main preparation work is the installation of Ubuntu system and ROS system.

[0129] The middle layer is a remote LAN built on the Pomegranate VPN software. It is the information interaction layer of the whole system. Without this channel, the master and slave cannot communicate, and control and data backhaul cannot be realized. It can be said to be the middle medium of the whole system, enabling the master and slave to interact at the Internet level. At the same time, it can also enable more users to use and interact, realizing the concept of Internet of Things. The main preparatory work is the installation of Pomegranate software and the configuration of ROS system master and slave.

[0130] The top layer is the decision-making layer, which can be either a human-in-the-loop system or a machine-controlled upper-level system. It is the central hub for issuing commands to the entire system. Without it, the vehicle chassis would not move, and feedback information could not be processed. It can be said to be the brain of the entire system and the core of the control algorithm. Through a control algorithm, the operating parameters of the vehicle chassis are further optimized, while providing a standard control method for autonomous driving. The main preparatory work is the installation of VM virtual machine and the installation and learning of MATLAB Simulink ROS toolbox.

[0131] Step 2: Design a multi-axle vehicle steering and torque distribution algorithm based on target speed control.

[0132] S2.1 Calculating the steering angle of a multi-axle vehicle using the Ackermann steering principle

[0133] Algorithm design is the core of six-axis chassis kinematic parameter optimization and the foundation of chassis kinematic analysis. Both normal maneuvering and future autonomous driving rely on kinematic algorithms. However, directly using C language for algorithm editing not only significantly increases the difficulty of writing the code but also makes it difficult for others to implement, resulting in extremely poor versatility and integration. Therefore, this embodiment chooses Simulink software as the basis for building the control algorithm. Simulink has numerous algorithm toolboxes and models, offering significant advantages for algorithm development and future applications. Specific steps include...

[0134] S2.11 Calculate the wheel steering angle

[0135] The Ackerman steering principle for electric vehicles refers to the ideal state in which, ignoring factors such as lateral forces and tire wear, the rotation center lines of all wheels of an electric vehicle can eventually intersect at a single point. The relationship between the inner and outer wheel turning angles of the front wheels, derived from the Ackerman principle, can be represented by a simplified geometric model. When four-wheel drive electric vehicles are designed based on the traditional Ackerman steering theory, they can ensure smooth turning at low speeds without tire wear, which not only improves tire lifespan but also increases the safety of driving electric vehicles.

[0136] like Figure 5 The left and right steering angles of the front wheels are shown. Let K be the center distance between the steering wheels and L be the wheelbase of the electric vehicle. The formula for the wheel steering angle is obtained by geometric calculation as follows:

[0137]

[0138] Where α1 is the steering angle of the left front wheel and α2 is the steering angle of the right front wheel;

[0139] Due to hardware limitations, the rotation angle of each wheel cannot be controlled separately; only the rotation angle of a single axle can be controlled. The difference in travel distance between the two wheels is resolved through external mechanical structures and torque differential control.

[0140] S2.12 Based on the calculated wheel steering angles of the multi-axle vehicle, calculate the relationship between the wheel steering angles and the Ackermann steering angle.

[0141] (1) As Figure 6 As shown, let angle a0 be a known angle, and the hypotenuse of the triangle be perpendicular to the direction of the wheel. Therefore, we can conclude that:

[0142] a = 90 - a0 (2)

[0143] f0 = 90 - f (3)

[0144] (2) Figure 7 As shown, given that the rotation angle of axle one is 'a', the distance of axle one from the midpoint is 'A', and the distance of axle two from the midpoint is 'B', let the distance of the midpoint from the turning center be 'S', and the rotation angle of axle two be 'b'. Given the distance of axle two from the midpoint is 'B', we can deduce that:

[0145]

[0146]

[0147]

[0148]

[0149]

[0150]

[0151]

[0152] (3) Figure 8 As shown, given that a0 is the program turning angle, side L is the wheel track (69cm), and side (A+B) is the distance from the axle to the turning point, we can deduce that:

[0153]

[0154]

[0155]

[0156] Therefore, given the turning angle of one axle and the distances between the midpoints of the turns of each axle, the turning angle of all six axles can be calculated. Here, in order to maximize the maximum turning angle of the six-axle vehicle, the first and last axles are selected as the maximum turning angles to constrain the steering control of the entire vehicle. This can be modified at will later through program control.

[0157] S2.2 Design a fuzzy torque control method based on the different speeds inside and outside the steering wheel.

[0158] Since the original chassis has two control methods, torque loop and speed loop, but the speed loop controls the speed of the entire vehicle and the steering differential is applicable to the original four-wheel chassis, it is not useful for the six-axle twelve-wheel chassis. Therefore, this embodiment selects the torque distribution algorithm of speed closed loop. The CAN bus protocol of the chassis has torque control frame 0x123 and real-time speed torque 0x177 and 0x178 corresponding to the CAN frames of the left and right wheels. In order to optimize the algorithm and better feedback the torque distribution effect under speed closed loop, this embodiment introduces Simulink in MATLAB software to optimize the closed loop distribution algorithm.

[0159] To establish the connection between Simulink and the ROS system, in order to enable remote control of the Logitech G29 steering wheel, this embodiment added a remote networking component to the entire system. This remote LAN channel can also be used to achieve remote control of the algorithm. The ROS1 module in the Simulink RosTool toolbox is used to build a ROS topic on Simulink.

[0160] The steering angle control of the entire vehicle has been set up. Now, it is necessary to design a torque distribution algorithm for multi-axle vehicles to adapt to the differential effect under the straight-line and turning conditions of multi-axle vehicles, so as to realize power distribution and reduce tire wear. In this embodiment, a fuzzy torque distribution algorithm based on speed loop calculation is established, which can realize the internal and external differential effect in a relatively simple and practical way, and also control the vehicle's motion information in a relatively quantitative way.

[0161] S2.21 Design Torque Distribution Control Algorithm

[0162] like Figure 9 As shown, the longitudinal vehicle motion is based on a PD-controlled full-vehicle speed control and torque distribution strategy, leveraging the advantage that multiple wheels are individually driven. A human-in-the-loop control is employed to control the target speed, and the algorithm calculates the torque distribution value to control the torque output of all 12 wheels of the vehicle. The torque distribution formula is as follows:

[0163] T=K P ×ΔV+K d ×(ΔV-V P )+T0 (14)

[0164]

[0165]

[0166] Because the car's steering is in an arc, the turning speed ratio is:

[0167]

[0168]

[0169]

[0170]

[0171]

[0172] angle=(Joy->axes[axis_ang]*angular) / 2;

[0173] S=174 / tan(angle*PI / 180.0f);

[0174] L = 69;

[0175] A1=((atan(104 / (S+69)+0.0f))*(-1))*2*180.0f / PI;

[0176] B1=(atan(23 / (S+69)+0.0f)*(-1)*180.0f / PI)*2;

[0177] By analogy, the steering angles of the multiple wheels turning outwards can be derived as follows:

[0178]

[0179]

[0180] Then: R1 = S / cos(a0);

[0181] R2 = (S + 69) / cos(A1);

[0182] V2 / V1 = R2^2 / R1^2

[0183] S2.22 Simulink Programming Implementation

[0184] like Figure 10 As shown, since the speed feedback of the vehicle chassis in this embodiment is fed back separately by the front and rear wheels, and the speed value is the average of the speeds of the front and rear wheels, it is necessary to process the feedback speed to achieve differential control of the left and right wheels. Let the feedback speed be V0, the left wheel speed be V2, and the right wheel speed be V1, and the vehicle is turning right, with the left wheel being the outermost wheel. Then...

[0185]

[0186]

[0187]

[0188]

[0189]

[0190]

[0191] a×V0=V1 (30)

[0192]

[0193] b×V0=V2 (32)

[0194]

[0195] Therefore, the algorithm for processing the real-time speed of the right wheel is V0*2 / (1+R2 / R1);

[0196] The algorithm for processing the real-time speed of the revolver is V0*2 / (1+R1 / R2);

[0197] Furthermore, when turning from right to left, the right wheel becomes the outer wheel, and the processing algorithm can handle the interaction between the left and right sides.

[0198] Step 3: Design the control algorithm module for controlling vehicle steering.

[0199] like Figure 10 The Simulink overall program diagram shown in this embodiment shows that the control algorithm is divided into three parts, each of which controls the torque of the four wheels of a vehicle.

[0200] S3.1 Design of S Calculation Module

[0201] like Figure 11 As shown, S is the vertical distance from the wheel steering center to the chassis, mainly used to calculate the turning angle of each wheel of a six-axle vehicle and the distance R of each wheel from the steering center; the algorithm has been derived in the previous text and will not be repeated here.

[0202] S3.2 Design Control System

[0203] like Figure 12 As shown, the first part of the control system corresponds to the first two axles of the six-axle chassis. It is mainly calculated by the distance between the four wheelbase steering centers. R1, R2, R3, and R4 correspond to the left front, right front, left rear, and right rear wheels of the vehicle, respectively. Specifically, it includes a target speed receiving module, four torque PD controllers, and a torque distribution and distribution module. Its specific design process includes:

[0204] S3.21 Target Velocity Receiver Module Design

[0205] like Figure 13 As shown, the target speed receiving module is used to receive the topic message of ECU0 and extract the target speed and shift message through the bus selector for reading; the reason for using three channels is that torque distribution of the six-axle chassis under steering conditions is to be performed later, and the control is divided into three lines: straight driving and crab steering, Ackermann right steering, and Ackermann left steering, which are selected and allocated by using the shift message to select the three lines.

[0206] S3.22 Torque PD Controller Design

[0207] (1) Design of the left front wheel control algorithm module

[0208] like Figure 14 The processing of the left front wheel control algorithm module involves selecting the PD algorithm and its input data. The angle output of the G29 steering wheel control program is used for judgment. When the angle is 0 degrees, it is determined to be either straight or crab steering (requires program control judgment), and shift = 0 is output. The switch module then judges and outputs the first input signal. When the angle is greater than 0 degrees, it is determined to be Ackermann right turn, and shift = 1 is output. The switch module then judges and outputs the second input signal. When the angle is less than 0 degrees, it is determined to be Ackermann left turn, and shift = 2 is output. The switch module then judges and outputs the third input signal.

[0209] exist Figure 14In the first path, the target speed message from the motor is directly read for use in the PD control algorithm, ensuring that all six axles and twelve wheels maintain the same target speed for output. The second path is the Ackerman right turn calculation algorithm. Because it is the front left wheel, it rolls on the outside during a right turn. Its calculation formula should be... R1 and R2 have already been calculated externally and input into this module via in4 and in5, and are used in the function; the third route is an Ackerman left turn, meaning this wheel is the inner wheel and the front wheel, i.e., the starting wheel. Therefore, the motor speed is directly assigned to the function and selected by the switch module for PD control. The Ackerman right turn algorithm module is as follows: Figure 15 As shown;

[0210] (2) Design of the left rear wheel control algorithm module

[0211] like Figure 16 As shown, in this module, the closed-loop negative feedback part of the PD algorithm in this embodiment uses the real-time vehicle speed information fed back by the vehicle. One vehicle feeds back two signals: the average speed of the two wheels of the front vehicle and the average speed of the two wheels of the rear vehicle. Therefore, when it is necessary to implement differential control of the two wheels, an algorithm must be used to decompose the real-time speed of the left and right wheels. That is, two function modules need to be designed in the feedback module. By selecting the shift signals for left and right turns and straight driving, the input is selected to the comparator for negative feedback. At the same time, the data processed by the PD algorithm is sent to the left front wheel of the vehicle through the torque distribution module and continuously iterates the feedback until the correction value of the target speed and the correction value of the real-time speed are equal, and it is always in a state of dynamic equilibrium.

[0212] The above control module is one axle of the six-axle vehicle and is also the starting axle for calculation in this embodiment. The speed of the four axles in the middle needs to be adjusted according to the speed of the starting axle. Therefore, R1 is needed to calculate the ratio of the speed of the subsequent wheels to the speed of the starting wheels in the calculation of the subsequent axles. Therefore, the out module is used in the module to transmit the parameters of R1 to the control modules of the other four axles for data calculation.

[0213] like Figure 17 As shown, in the left rear wheel control algorithm module, there is a left rear wheel allocation algorithm for a vehicle. This algorithm is the correction part of the left rear wheel algorithm. The calculation result of R1 is imported into the module through the in6 module. Since the left turn is the inside wheel, its reference speed should be the speed of the front left wheel, i.e., the motor speed, plus the ratio.

[0214] like Figure 18 As shown, the reference speed of the outer wheel is no longer the motor, but the speed of the inner wheel, i.e., the target speed calculated by the algorithm for the left inner wheel, and the output speed of port y.

[0215] (3) Module torque distribution module design

[0216] like Figure 19 As shown, the torque distribution module uses the pub function module and empty message module of the ROS toolbox to send messages. The compiled custom message is imported into MATLAB. In the function interface of the empty message module, the torque distribution information frame, ecu_123, is selected. The bus distributor module decomposes the message input from the bus into a data empty frame, and the four input ports in1, in2, in3, and in4 transmit the torque data into the message frame. Then, the pub module publishes the formatted ecu_123 to the torque topic / ecu1 of the vehicle to realize the torque distribution control.

[0217] The six-axis unmanned vehicle chassis based on the ROS general robot operating system established by the above method includes

[0218] Example 2: To verify the effectiveness of the six-axis unmanned vehicle chassis based on the ROS general robot operating system designed as described in Example 1 above, this example is designed to conduct a multi-axis vehicle torque coordination distribution experiment and analysis on the multi-axis unmanned vehicle chassis based on the ROS general robot operating system described above:

[0219] 1. Bench Test Analysis under P and PD Controllers

[0220] Because real-vehicle testing involves great uncertainty and danger, I chose to conduct preliminary experiments based on a suspended test bench under laboratory conditions before real-vehicle scenario testing. This was to examine the control coupling between the algorithm and the actual hardware, and to modify the software algorithm and parameters based on specific hardware parameters to better improve the hardware performance.

[0221] 1.1. Test Results of the P Controller on the Test Bench

[0222] like Figure 20 As shown, the simulation data is the torque output data obtained by inputting a target speed of 0 and a real-time speed of 5 revolutions per minute based on the mathematical model. The P controller is used here. It can be seen that the coefficient is set too large. This is mainly because the torque output of the physical model is affected by the motor. It cannot rotate at a smaller torque value, so a larger coefficient is set.

[0223] like Figure 21 As shown, by inputting a real-time speed of 5 revolutions per minute and a steering angle of 60 degrees, we can check whether the real-time speed differential correction algorithm works properly during steering. It can be seen that under the control information of a 60-degree left turn, the speed correction of the left wheel decreases and the speed of the right wheel increases. Of course, since the target speed is not input and there is no actual vehicle response, we cannot see the data correction during the increase.

[0224] like Figure 22 As shown, this is the experimental data from the P controller debugging conducted using a real vehicle chassis test. It can be seen that in the initial stage, the real-time speed reception was poor, indicating some issues with real-time operation. After a period of time, the system began to work normally. Upon inputting a target speed, due to the performance characteristics of the vehicle's motor, a large torque is required for startup, resulting in a real-time speed of 0. The torque distribution signal (yellow line) shows a large output, which then quickly drops back. While the real-time speed and the target speed are close but not overlapping, this is caused by the P controller's parameters being set too low and the presence of static errors. This also demonstrates a significant difference between the actual physical model and the digital analysis model. Future parameter design should consider the real-world physical influences. The parameters are adjusted to bring the control law values ​​closer to larger values ​​to ensure better control performance. Meanwhile, it can be seen that when the target speed is reduced to 0, i.e., when the accelerator is released, the torque distribution output shows a large negative value. This is because the real-time speed still exists, and a negative deviation occurs between the target speed and the real-time speed. After P control calculation, a negative torque value is output, effectively braking the vehicle. This differs significantly from actual human driving, primarily due to the safety and reliability requirements of autonomous driving. However, when the control signal disappears, the autonomous vehicle must return to a stationary state and cannot move. Furthermore, the input speed of 0 is also a control signal for the autonomous vehicle; it cannot be set to the rolling state of a human driver.

[0225] 1.2. Test Results of PD Controller on Test Bench

[0226] like Figure 23 As shown, the control algorithm for the last two wheels here uses the PD algorithm, while the first two wheels use P control. By comparison, it can be intuitively seen that under PD control, there is a large jump phase in the initial stage, which is the effect of using D. It can approach the target speed faster, but it will produce a large overshoot. However, it is precisely because of the existence of its overshoot that it is more suitable for the control effect of our actual physical model. Therefore, theoretically, using PD control here should be able to achieve a better torque distribution effect and improve the real-time response performance of the vehicle.

[0227] like Figure 23 As shown, this is the effect of a real vehicle bench test using a PD controller. The test results show that after adding the PD controller, the torque output response is very fast, but the degree of fluctuation is also increased, and the instability of the system changes significantly. At the same time, due to the setting of the PD controller and the underlying issues of real vehicle real-time speed interpretation, the PD controller still cannot track the target speed well, and parameter adjustment is required in the field.

[0228] 2. Real-vehicle scenario experiment under P controller

[0229] After preliminary bench tests and analysis of the experimental data and results, the control algorithm has basically achieved the various control requirements, and the vehicle driving safety has been guaranteed to a certain extent. It is now possible to conduct outdoor tests on real vehicles and adjust the PD control parameters of the vehicles under actual operating conditions to further optimize the actual vehicle operation effect.

[0230] 2.1. Road Experiment Analysis under P Controller

[0231] like Figure 24 As shown, this is the result of a real-vehicle road test conducted on a closed campus road. This graph clearly shows the torque required for the tires to overcome static friction and the resistance of the motor during vehicle startup. It also shows the friction force exerted on the tires by the road under the vehicle's significant weight during actual operation, which is significantly lower than the torque output during the bench test—nearly three times the difference. This part of the experiment includes steering; the changes in the orange line reflect the differential control characteristics implemented by the control algorithm during vehicle steering. The target speed distribution to the left and right wheels has basically met the algorithm's control requirements, but the real-time speed correction algorithm still has some issues with the real-time speed correction of the left and right wheels. During steering, the real-time speed correction ratio is not as... The positive results are evident in the 40-50 second timeframe shown in the graph. It's clear that the algorithm effectively corrected the real-time speed of the left and right wheels. A significant difference also lies in the real-time speed measurement. Due to the complexity of real-world road conditions, with bumps and potholes, external interference is more pronounced, resulting in a noticeable fluctuation in the real-time speed (blue line) compared to the bench test. A comparison of the top and bottom reveals that the torque output of the front wheels is significantly lower than that of the rear wheels. This is a characteristic of real-world road driving. Firstly, the main load of a single vehicle, namely the battery, is located in the rear-wheel drive. Secondly, the latency of the algorithm and control may cause the front wheels to obstruct the rear wheels, leading to a difference in torque output between the front and rear wheels. This aligns with the theoretical analysis and actual performance.

[0232] 2.2. Road Experiment Analysis under PD Controller

[0233] like Figure 25As shown, the PD controller is more susceptible to external interference under actual road conditions, especially when the D parameter is set to 0.1. However, the torque distribution is very large, mainly due to the short control cycle, which makes the derivative term more susceptible to oscillations caused by interference. In some cases, the derivative term may even reach infinity, causing the program to crash and stop running. However, it is also clear that the speed output is faster and more accurate than that under the P controller, and the torque response is very rapid with a large output value. Of course, for the problems that occurred in the experiment, this embodiment will start with the PD control value and constrain the calculation parameters of other terms to ensure the finiteness of the derivative result term, thereby improving the reliability of the system.

[0234] Through the above experiments and theoretical analysis, it was found that the torque distribution algorithm under the speed loop can effectively distribute the torque of the twelve wheels of the system, ensuring the basic stability of the target speed. It realizes the fuzzy self-allocation strategy of torque, has good practical value, and can be used as the algorithm basis for realizing autonomous driving.

[0235] This ROS robot system has good development potential. The six-axis chassis built on the open-source ROS platform can serve as a relatively complete experimental platform to verify the motion algorithm of multi-axis special vehicles. It can reflect the motion characteristics of multi-axis vehicles well and has certain significance for experimental applications and teaching. More in-depth theoretical exploration and engineering practice can be carried out based on this platform.

[0236] Meanwhile, this six-axle chassis uses a modular construction method, allowing for future modifications to the wheelbase and number of axles. It can be used for algorithm design and experimental analysis of more vehicle models. Furthermore, due to the largely consistent chassis parameters, it can also utilize parallel driving functionality to create a simulated driving system, providing a simple platform for driver training of heavy-duty multi-axle special vehicles. The above describes and illustrates the basic principles, main features, and advantages of this invention. Those skilled in the art should understand that this invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made without departing from the spirit and scope of the invention, and all such changes and modifications fall within the scope of the invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A design method for a multi-axis unmanned vehicle chassis based on the ROS general robot operating system, characterized by: include Step 1: Design the hardware platform for multi-axle vehicle electric drive S1.1 Hardware Architecture Design; S1.2 software platform design; Step 2: Design a multi-axle vehicle steering and torque distribution algorithm based on target speed control. S2.1 Calculates the wheel steering angle of a multi-axle vehicle using the Ackermann steering principle. The calculation process includes: S2.11 Calculate the wheel steering angle Let K be the center distance between the steering wheels and L be the wheelbase of the electric vehicle. The formula for the wheel steering angle, derived geometrically, is as follows: （1） Where α1 is the steering angle of the left front wheel and α2 is the steering angle of the right front wheel; S2.12 Based on the calculated steering angle of the multi-axle vehicle wheels, calculate the relationship between the wheel steering angle and the Ackermann steering angle; The calculation process for the relationship between the wheel angle and the Ackermann angle includes: (1) Let angle a0 be a known angle. The hypotenuse of the triangle is perpendicular to the direction of the wheel. We can then conclude that: （2） （3） (2) Given that the rotation angle of the first axle is a, the distance of the first axle from the midpoint is A, and the distance of the second axle from the midpoint is B, let the distance of the midpoint from the turning center be S, and the rotation angle of the second axle be b. Given that the distance of the second axle from the midpoint is B, then we can deduce that: （4） （5） （6） （7） （8） （9） （10）； (3) Given that a0 is the program turning angle, L is the wheel track, and A+B is the distance from the axle to the turning point, we can deduce: （11） （12） （13） Therefore, given the rotation angle of one axle and the distances of each axle from the midpoint of the turn, the rotation angles of multiple axles can be calculated. S2.2 A fuzzy torque control method is designed based on the difference in steering speeds inside and outside the steering wheel. The specific process includes: S2.21 Design torque distribution control algorithm; S2.22 processes the feedback speed to achieve differential control of the left and right wheels; The design process of the torque distribution control algorithm includes: The longitudinal vehicle motion is based on a PD-controlled overall vehicle speed control and torque distribution strategy, leveraging the advantage that multiple wheels are individually driven. A human-in-the-loop control is employed to manage the target speed, and the algorithm calculates the torque distribution value to control the torque output of multiple wheels throughout the vehicle. The torque distribution formula is as follows: （14） （15） （16） Because the car's steering is in an arc, the turning speed ratio is: （17） （18） （19） （20） （21） By analogy, the steering angles of the multiple wheels turning outwards can be derived as follows: （22） （23） Step 3: Design the control algorithm module for controlling vehicle steering.

2. The design method for a multi-axis unmanned vehicle chassis based on the ROS general robot operating system according to claim 1, characterized in that: The hardware architecture design includes (1) A three-level embedded hardware platform is used, and different processors are used in each level. The Nano processor is used as the master controller of the multi-axis homing chassis, and the PC is used as the slave. (2) The whole system combines multiple bus structures such as local area network, USB interface, and CAN bus. The PC and the underlying Jetson Nano processor use the Dandelion remote networking software to build a local area network under the Internet and exchange data and remote control through the TCP / IP protocol.

3. The design method for a multi-axis unmanned vehicle chassis based on the ROS general robot operating system according to claim 1, characterized in that: The software platform design process includes The lowest-level ROS chassis control software is a software system used to directly control vehicle chassis information and acquire chassis information; The middle layer is a remote local area network built on the Dandelion VPN software, which is the information interaction layer of the entire system; The top layer is the decision-making layer, which is the host computer layer where humans are in the loop or the machine controls itself. It is the central hub for issuing instructions to the entire system.

4. The design method for a multi-axis unmanned vehicle chassis based on the ROS general robot operating system according to claim 1, characterized in that: The process of processing the feedback speed includes Let the feedback speed be V0, the left wheel speed be V2, the right wheel speed be V1, and the car be turning right with the left wheel being the outermost wheel. Then we have: （24） （25） （26） （27） （28） （29） （30） （31） （32） （33） Therefore, the algorithm for processing the real-time speed of the right wheel is V0*2 / (1+R2 / R1); The algorithm for processing the real-time speed of the revolver is V0*2 / (1+R1 / R2); Furthermore, when turning from right to left, the right wheel becomes the outer wheel, and the processing algorithm can handle the interaction between the left and right sides.

5. The design method for a multi-axis unmanned vehicle chassis based on the ROS general robot operating system according to claim 1, characterized in that: The design process of the control algorithm module includes: S3.1 Design S calculation module; S3.2 Design Control System S3.21 Design target speed receiving module; S3.22 designs a torque PD controller.

6. The design method for a multi-axis unmanned vehicle chassis based on the ROS general robot operating system according to claim 5, characterized in that: The design process of the torque PD controller includes (1) Design the left front wheel control algorithm module; (2) Design the left rear wheel control algorithm module; (3) Design module torque distribution module.

Images