8. Guide to Motion Control Simulation

8. Guide to Motion Control Simulation

8.1 Simulation Description

We provide a mujoco-based simulation system for the A2 robot, designed to serve as a testing ground for basic calls to the A2 robot’s motion control and action playback interfaces. This serves as a basic validation before actual operation, reducing the risk of damaging the robot while facilitating the debugging of secondary development program logic.

The system runs the mujoco_sim, motion_control, and motion_player components. mujoco_sim acts as the underlying simulation engine, simulating the robot’s physical body and hal_ethercat. motion_control interacts with mujoco_sim to exchange control commands and status, thereby controlling the simulated robot. motion_player, on the other hand, calls the high-level interfaces provided by motion_control to play actions. The motion_control and motion_player components are identical to those on the real robot, allowing the system to call all interfaces of the motion control and action playback modules.

8.2 Prerequisites

This simulation system is provided as a container image, containing executable files for each component, action resource files, and calling scripts. To use this simulation system, the following conditions must be met:

1. An x86_64 architecture Linux system with Docker installed.

2. An X11 desktop system.

   • You can check this by running echo $XDG_SESSION_TYPE, and the expected output should be x11.

8.3 Usage Tutorial

8.3.1 Pulling the Image

1. Log in to the read-only account of the image repository

docker login tongyong-public-cn-shanghai.cr.volces.com -u crrobot@aima-public-reader -p Aima123456

2. Pull the corresponding image

docker pull tongyong-public-cn-shanghai.cr.volces.com/aima-public/a2-lite-simulator:v2.0

8.3.2 Starting the Image

1. Allow Docker to access the display, which is used to configure the access control list for the X Window System. Running xhost + will allow all hosts to connect to the current user’s X server, effectively disabling access control for the X server and allowing any user to access and operate it.

   xhost +

2. Start the Docker container

   docker run -it \
   --name=a2-lite-sim \
   --privileged \
   --net=host \
   --ipc=host \
   --pid=host \
   -e DISPLAY=$DISPLAY \
   -v /tmp/.X11-unix:/tmp/.X11-unix \
   -v /run/dbus/system_bus_socket:/run/dbus/system_bus_socket:ro \
   -d tongyong-public-cn-shanghai.cr.volces.com/aima-public/a2-lite-simulator:v2.0

   If you find that the simulation interface runs sluggishly on a machine with an NVIDIA graphics card after starting the container with the above command, try using the following command to start the container:

   docker run -it \
           --name=a2-lite-sim \
           --gpus all \
           --privileged \
           --net=host \
           --ipc=host \
           --pid=host \
           -e DISPLAY=$DISPLAY \
           -e NVIDIA_VISIBLE_DEVICES=all \
           -e NVIDIA_DRIVER_CAPABILITIES=all \
           -v /tmp/.X11-unix:/tmp/.X11-unix \
           -v /run/dbus/system_bus_socket:/run/dbus/system_bus_socket:ro \
           -d tongyong-public-cn-shanghai.cr.volces.com/aima-public/a2-lite-simulator:v2.0

3. Enter the container

   docker start a2-lite-sim && docker exec -it a2-lite-sim /bin/bash

8.3.3 Starting the Simulation, Motion Control, and Action Playback Modules

1. After entering the image, run the following command to start the simulation

   docker start a2-lite-sim && docker exec -it a2-lite-sim /bin/bash
   
   cd /home/agi/a2_simulation/mujoco_sim/bin
   ./start_a2_lite.sh

2. In a new window, connect to the container and start the motion control program

   docker start a2-lite-sim && docker exec -it a2-lite-sim /bin/bash
   
   cd /home/agi/a2_simulation/motion_control_lite_2.0.10/scripts/motion_control
   ./start_motion_control.sh

3. In another new window, connect to the container and start the action playback program

   docker start a2-lite-sim && docker exec -it a2-lite-sim /bin/bash
   
   cd /home/agi/a2_simulation/motion_player_lite_2.0.10/scripts/motion_player
   ./start_motion_player.sh

8.3.4 Controlling Robot Walking

Here, we will use robot walking as an example to explain the basic logic of calling a motion control interface. For other interfaces, refer to the motion control interface documentation and example scripts.

1. Open a new window, enter the container, and switch the motion control Action to the joint standing mode (RL_JOINT_DEFAULT)

   cd /home/agi/a2_simulation/tools-lite/motion_control
   ./SetAction.py # Select the number corresponding to RL_JOINT_DEFAULT

2. Click the Load-Key button in the mujoco_sim interface to stabilize the robot

   

3. Continue to switch the motion control Action to the enhanced walking mode (RL_LOCOMOTION_DEFAULT)

   cd /home/agi/a2_simulation/tools-lite/motion_control
   ./SetAction.py # Select the number corresponding to RL_LOCOMOTION_DEFAULT

4. Call the walking example script

   source /home/agi/a2_simulation/prebuilt/ros2_plugin_proto_x86_64/share/ros2_plugin_proto/local_setup.bash
   ./walk.py

8.3.5 Controlling Robot to Perform Actions

1. Switch the motion control state to the enhanced walking upper limb servo mode (RL_LOCOMOTION_ARM_EXT_JOINT_SERVO)

   cd /home/agi/a2_simulation/tools-lite/motion_control
   ./SetAction.py # Select the number corresponding to RL_LOCOMOTION_ARM_EXT_JOINT_SERVO

2. Call the SendMotionCommand interface to perform the action

   cd /home/agi/a2_simulation/tools-lite/motion_player
   ./send_motion_id.sh /agibot/data/resources/default/motion/motion_player/default/双手挥手/双手挥手