“In recent years, various types of military robots have played an important role in the field of national defense and geostrategy. In the vast western region of my country, due to the complex terrain and rugged roads, traditional wheeled or tracked robots cannot meet the requirements of terrain passability, while bionic quadruped robots can better meet the task requirements of reliable travel in unstructured terrain. . The hydraulically driven bionic quadruped robot is a research hotspot and main project at home and abroad in recent years. Among its key technology groups, electro-hydraulic servo control technology is the core technology to ensure the stable running function of the bionic hydraulic quadruped robot.
In recent years, various types of military robots have played an important role in the field of national defense and geostrategy. In the vast western region of my country, due to the complex terrain and rugged roads, traditional wheeled or tracked robots cannot meet the requirements of terrain passability, while bionic quadruped robots can better meet the task requirements of reliable travel in unstructured terrain. . The hydraulically driven bionic quadruped robot is a research hotspot and main project at home and abroad in recent years. Among its key technology groups, electro-hydraulic servo control technology is the core technology to ensure the stable running function of the bionic hydraulic quadruped robot.
1 Overall Design
1.1 Analysis of Control Objects
This paper relies on a bionic hydraulic quadruped robot being developed by the Beijing Institute of Technology Special Robot Technology Innovation Center. Each leg of the robot has 3 active degrees of freedom and 1 passive degree of freedom, which are the hip lateral swing joint and the hip orthodontic swing respectively. Joint, knee and foot second-order spring shock absorbers, all 12 active joints are driven by hydraulic cylinders. The structural dimensions of the robot are 120 cm long and 50 cm wide, the equivalent length of the thigh section rod (the distance from the hip joint axis to the knee joint axis) is 40 cm, and the equivalent length of the calf section rod (the knee joint axis to the foot end) is 40 cm. Envelope center distance) is 40 cm. The self-weight is 118.5 kg under the experimental conditions without integrated on-board hydraulic oil source. The overall structure of the quadruped bionic robot is shown in Figure 1.
1.2 System Architecture Design
Considering the servo control tasks comprehensively, it can be found that there are several contradictions in the system performance requirements, such as the time resource conflict between multi-channel parallel servo control and single-channel servo control accuracy and the hardware resource conflict of multi-input/output interface design. It is difficult for a centralized control system using a single controller to achieve a balance in the allocation of time resources and hardware resources, thus failing to meet the task index requirements. Based on the above goals, the system adopts a distributed system architecture to design the electro-hydraulic servo control system. In this system, four servo controllers are designed under the motion controller, as shown in Figure 2, and are connected through the servo bus to form a distributed electro-hydraulic servo control system.
The 4 servo controllers are respectively responsible for the servo control of the 3 hydraulic servo units on one leg of the robot. Through the distributed system architecture design, on the one hand, the splitting simplifies the control tasks, making the task sequence of a single controller more regular, and improving the stability of the software system; on the other hand, it realizes the corresponding control system and mechanical structure. Modular design. The design takes into account the requirements of precision and real-time performance in system tasks.
2 Servo bus interface design
The servo bus interface is the bus interface between the motion controller and the four servo controllers. It is responsible for transmitting the servo commands of the downstream hydraulic servo units and feeding back the working status of the upstream hydraulic servo units. It needs to have network characteristics. Commonly used buses mainly include RS422/485 serial communication bus, Ethernet, I2C bus, SPI bus and CAN bus. The CAN bus topology is flexible and changeable, and has no master-slave characteristics. Any working node on the network can send data at any time to realize point-to-point and point-to-multipoint data transmission. Using non-destructive arbitration technology, mailbox and ID determine the priority of node data. The smaller the ID, the higher the priority.When the short frame structure is used for transmission, each frame has 8 valid bytes, the transmission time is short, and the anti-interference ability is strong.. In this design, CAN bus is selected as the servo bus of the system, and the topology is shown in Figure 3.
The update frequency of the servo bus is 200 Hz, and the length of the servo command frame is 6 bytes per axis. A single frame servo command can be designed using the CAN protocol. The total length of a single cycle command is 72 bytes. The throughput calculation per cycle is as follows:
Among them, bpf is the bit per frame, f is the frame, the bus baud rate can be calculated to be 374.4 kbps, and the CAN bus communication baud rate can reach 1 Mbps within 40 m, which can meet the bus communication rate requirements. Considering the high real-time requirements of servo commands, it is necessary to avoid the problem that feedback information packets occupy the bus and affect the data timeliness. Therefore, a two-wire design is used in the design, and CAN1 is only used for the motion controller to issue command data to the servo controller. package, CAN2 is used for servo control to return sensory feedback information to the motion controller. The schematic diagram of the CAN transceiver circuit is shown in Figure 4.
Select the SN65HVD232 CAN bus transceiver. In Figure 4, R1 is the terminal load resistance of the CAN bus, and it is determined whether to short the jumper JP1 according to the bus position of the servo controller. When the corresponding servo controller is located at the end of the bus, in order to prevent signal reflection, short the jumper JP1, and pass the termination load resistance into the differential signal loop to suppress signal reflection interference. The CAN bus controller uses the on-chip bxCAN of the STM32F4 series MCU, and the two bxCAN peripherals control the command bus and the sensor feedback bus respectively. The bus drive design of the servo controller is shown in Figure 5.
3 Servo valve control interface design
Commonly used output stage designs include analog direct drive, DAC power amplifier semi-digital drive, digital power drive. This paper uses PWM to drive the MOSFET full bridge to achieve digital power drive, and uses MOSFET SI4405P and SI4404N to build a power drive full bridge.The STM32F405 internal timer is used to generate complementary PWM to drive the corresponding bridge arm. This control method first realizes direct digital control without using analog devices; secondly, it uses a single power supply to achieve bipolar control, which simplifies power supply and circuit design, and can be used. The superposition of the flutter signal in the servo valve drive signal is realized by adjusting the PWM carrier frequency.
4 Sensor feedback interface design
Both the position sensor and the pressure sensor used in the hydraulic drive unit of the robot use the form of voltage signal output. The STM32F405 integrates two 16-channel 12-bit high-performance internal ADCs. The single-channel sampling frequency can reach 2.4 Msps, and the 6-channel polling sampling is the fastest. Guaranteed 400 ksps per channel, can meet the 1 kHz servo frequency and accuracy requirements, so this design does not use an independent external ADC, and directly uses the internal peripherals. The sensor output signal is 0~10 V, and the conditioning circuit is shown in Figure 6.Figure 6 Schematic diagram of passive gain sensing signal conditioning circuit
The ADC uses the STM32F405 internal peripherals, the single-channel sampling frequency is set to 16 ksps, and the 16-byte drum buffer is written through DMA for digital filtering. The software flow is shown in Figure 7.
5 Servo bus command protocol design
Configure the CAN bus communication ID as a standard 11-bit ID. By configuring the 11-bit ID, the designed servo command ID format is as follows.
The ID0 and ID1 bits are the number coding bits of the servo controller, which are coded in the order of the left front leg No. 0, the right front leg No. 1, the left rear leg No. 2, and the right rear leg No. 3. The ID2 and ID3 bits are the joint number coding bits, which are 0, 1 and 2 respectively for the top-down hip swing, hip swing and knee joint of a single leg. The specific codes are listed in Table 1.
The ID5~9 bits are used for instruction encoding. Because the servo bus and the feedback bus are independent of each other, the servo control and status feedback are designed separately. The ID4 bit is the broadcast flag bit. When this bit is 1, the corresponding command is sent to each channel in the form of broadcast. The ID10 bit is the read and write flag bit of the servo bus. When this position is 0, the control command represents the read command for the corresponding control quantity, and the servo controller sends back packet data through CAN2.
5.1 Servo control bus command protocol
The servo control command is issued by the motion controller through CAN1, and is designed according to the response mechanism of the CAN bus to the ID. The high-priority command has a smaller control word and a higher bus arbitration priority. Design control instructions omitted – editor’s note.
5.2 Status feedback bus command protocol
Status feedback command is used to feed back self-check information and return working status through CAN2 – Editor’s Note.
6 Design of Fuzzy Adaptive PID Algorithm for Asymmetric Feedforward Compensation
The motion of the bionic hydraulic quadruped robot is driven by the hydraulic actuator system, and the robot itself has the characteristics of variable load and time-varying environment, so the control object is a nonlinear and time-varying system.Fuzzy PID control has great advantages in this kind of system control and can improve the control performance.
In order to reduce the complexity of the fuzzy controller, this system uses an error segmented intelligent control algorithm, and the principle of its control scheme is shown in Figure 8. The identification switch in the figure selects BangBang control or fuzzy adaptive PID control by judging the error threshold. When the error is greater than the set threshold, BangBang control is performed; when the error is less than the set threshold, fuzzy adaptive PID control is performed.The parameters of the fuzzy reasoning expert library are obtained through the self-tuning system in the debugging mode.
Aiming at the asymmetry and hysteresis characteristics of the hydraulic control system, the speed and acceleration feedforward control parameters are added on the basis of the above controller design to compensate the hysteresis of the system and improve the response speed. By setting different feedforward parameters for the two motion directions, asymmetric compensation for the motion directions of the rod cavity and the rodless cavity is realized. The specific implementation method is that the compensation parameters of the two motion directions are selected by one direction judgment switch, so as to realize the control. At the same time, the asymmetry of the hydraulic cylinder control parameters is reflected in the fuzzy strategy table.
One end of the hydraulic cylinder was fixed on the experimental platform, the other end was empty or a 10 kg load was connected in series, and a 21 MPa high-power external oil source with power redundancy was used to supply oil.
7.1 Sinusoidal Position Control Experiment
For 85 mm stroke hydraulic servo system, the response results obtained by inputting 20 mm amplitude and 4.5 Hz sine wave excitation are omitted – editor’s note. It shows the servo control effect of symmetrical performance and good tracking of position and speed.
7.2 Step Response Experiment
A square wave excitation with an amplitude of 30 mm was applied to the hydraulic servo system with a stroke of 100 mm of the knee joint to realize a step signal with a stroke of 60 mm in both forward and reverse directions. The steady-state error of 0.01mm can be achieved within three oscillations, and the experimental response curve is abbreviated – Editor’s Note. Recording experimental data table omitted – editor’s note.
This paper designs a distributed electro-hydraulic servo control architecture based on STM32F405, builds a servo bus using CAN bus, and designs the electro-hydraulic servo input and output interfaces in a targeted manner. Fuzzy adaptive with asymmetric feedforward compensation is introduced. control algorithm. After experimental verification, the servo control level of the controller meets the performance index requirements and achieves a good control effect.