【Training Camp】Bionic Quadruped Robot Dog

Project Introduction

Demo video of Station B : [Lichuang Training Camp] Self-made Hongmeng robot dog, open source software and hardware

Open Source Link : Bionic Quadruped Robot Dog

A four-legged robot dog based on Chuanzhi Hi3861 module. The legs are composed of a total of 8 mg90s servos, each leg has two degrees of freedom. Based on the inverse solution of kinematics, it can realize rich action modes. Bluetooth remote control and preliminary verification of face following functions have been added.

References

• Video tutorial from Station B: https://www.bilibili.com/video/BV1RU4y1D75p?share_source=copy_web
  
• Sample project open source link: ohosdog - Jialichuang EDA open source hardware platform (oshwhub.com)
  
• Hongmeng development information: Hongmeng openharmony software and hardware development tutorial_bilibili_bilibili
  
  HUAWEI DevEco Device Tool User Guide-OpenHarmony Device Development-HarmonyOS Device Development
  
• Main control board information: Chuanzhi Hongmeng Yuanqi Pai (czxy.com)
  

circuit design

Main component selection

• Main control: Hi3861 module
  

• Servo drive IC: pca9685
  

• Power supply: 7.4V 2s battery power supply, with 3P charging interface
  
• Power management: 5v step-down dcdc or LDO (LM1084) + AMS1117-3.3
  
• IO interface: 16 servo interfaces
  
• Expand IO: mpu6050+ultrasonic sensor+Bluetooth serial port
  
• TypeC+CH340 serial port download circuit
  
• switch
  
• Power Indicator
  
• Diode power isolation
  
• Servo: mg90s 180 degrees, 8 pcs
  

Connection pin assignment
Servo drive IIC
GPIO10 --> SDA
GPIO9 --> SCL

ultrasonic sensor

GPIO07 -->trig

GPIO08 -->echo



All the remaining IOs are led out, and the 3.3V and 5V power interfaces are led out at the same time to facilitate the power supply of external modules.

PCB design

• 3D preview
  
  
  

Design files can be found in Lichuang EDA project files.

structural design

• Drawing reference: Solidworks - Robot Research Institute (heimarobot.cn)
  

See attachment for solidworks modeling files硬件部分/hardware/structure

Structural optimization

• The volume of the sole of the foot is expanded to increase friction.
  
• Added side and bottom slots to reduce overall weight and facilitate expansion module installation
  

Fastener selection

• M2 screw nut
  
• M2.5 10 nut
  
• M2.5 16 screw nut
  
• M3 15 double-pass copper pillar
  
• M3 40 double-pass copper pillar
  
• M3 nut
  

3D printing

Use the light-curing printing provided by Jialichuang 3D Monkey, the material is SLA photosensitive resin

software design

motion planning

Parameter calibration

• Servo position record
  
  
  
• Servo number definition
  
  The above picture is the position number provided in the tutorial. Since the video was not referenced when designing the circuit, the actual connection method is inconsistent with the picture. The corresponding number is given according to the following definition:
  

//leg redefine
// left front leg
#defineLF_A 12
#defineLF_B 13
// left back leg
#defineLB_A 2
#defineLB_B 3
// right front leg
#defineRF_A 11
#defineRF_B 10
// right back leg
#defineRB_A 5
#define RB_B 4

• Leg parameter length measurements (to the right is the positive direction, with the right front leg as the standard)
  
  The actual leg length measurement after installation is inconsistent with the software default length and needs to be modified based on the actual situation.
  
  
  

Inverse solution algorithm verification

The inverse solution code is provided in the open source project file, and the core function is the dog_leg_move() function. It is necessary to modify the corresponding parameters according to the actual design situation. Extract the inverse solution operation code and run the test in dec c++. According to the actual test results, adjust the algorithm. On the one hand, the alpha and beta positions need to be exchanged, and the beta angle output differs by 180 degrees. On the other hand, the angle output result must be modified. , the final verification situation is as shown below, and the inverse solution result is very accurate. The modified code has been placed in the attachment and can be viewed by yourself.

gait design

• step on
  
  The designed motion position coordinates are (0, 60) --> (0, 70), set a certain periodic cyclic motion, and the stepping speed is determined by this period. Design a simple gait: move diagonally at the same time, up and down on the same side.
  
• locomotion gait
  
  Use the gait simulator to set up and export the two-dimensional array of the end trajectory, and execute it in a loop. Because what you get is coordinate data, you need to choose to export path trajectory data instead of path point data.
  
• Forward and backward: diagonal movement is performed simultaneously, and backward movement is in the opposite direction.
  
• Left and right steering: The movement method is the same as the front and rear direction, but the left and right gait distance is adjusted, similar to differential steering.
  
• Entertainment action design:
  
• wave
  
• (pee...)
  

Hongmeng Development Environment

• Main compilation directives
  

hb set //Initialize the development environment and operate the source code once
hb clean //Clear compilation
hb build //compile

• Development Notes
  
  Based on Hongmeng lightweight system kernel LiteOS-M, it can adopt development methods under various subsystems and real-time operating system frameworks.
  
  A task thread was created during development, and all functions are implemented in this thread. For usage, please refer to the main control board information provided at the beginning of the article.
  

Hi3861 serial port driver development

• Use serial port 1 and connect:
  

GPIO6--UART1_TX
GPIO5--UART1_RX

K210 identification and following

• Reference: Introduction to MaixPy documentation - Sipeed Wiki
  

Run the face detection model

The face detection model is based on the yoloV2 framework. After burning the model and running the recognition program, the measured detection frame rate can reach 10+fps, and the effect is remarkable.

Reference: [K210 Study Notes] #MaixPy# (3) Face detection and face recognition

• Serial communication program
  
  Connect
  

GPIO8--TX
GPIO21--RX

k210--Hi3861
  08--05
  21--06
  07--14
  15--11

• IO communication mode
  
  Control port p15, 3 IO ports: p07 p21 p08, a total of eight identification statuses can be transmitted
  

ESP32 Bluetooth remote control program

• ESP32 program
  
  Developed based on the Arduino environment, it receives control instructions sent by the mobile phone and passes them to the main control chip through the IO port control.
  
• The mobile phone connects to the ESP32 via Bluetooth and sends control instructions with the help of a customized Bluetooth assistant.
  
  
  

Submit code instructions

• Main control code
  
  The Hi3861 part is developed using genkipi open source firmware. The following two files are modified (note the file path). See the attachment `software part/software/final_code/app`
  
  
  
• ESP32 code
  
  see attached软件部分/software/esp32_control
  
• K210 code
  
  see attached软件部分/software/k210
  
• Can directly burn firmware
  
  A total of 11 versions of the debugger firmware are recorded, see软件部分/software/bin
  
• Inverse kinematics solution test code
  
  Compile using devcpp environment and run directly, see attachment软件部分/software/inverse_test