⚙ Development Guide > SDK > Concept Guide

Control Modes

Control modes are built into the robotic arm to accommodate different usage scenarios. Select the appropriate mode for your scenario from the table below to obtain better control results.

Control Target	API Method	VLA Model Inference (direct_control)	Teleoperation (servo_control)	Autonomous Grasping (with Planning) (planning_control)	Force Control (Compliance & Admittance) / RL Inference (mit_control)
Robotic arm joints	move_joint	✔	✔	✔	✔
Robotic arm tool end	move_end_pose	✔	✔	✔	✔
	move_end_pose_linear	✘	✘	✔	✘
	move_end_pose_circle	✘	✘	✔	✘
End effector	move_eef	✔	✔	✘	✔

Control-mode descriptions:

• direct_control (direct motor position and velocity control with low latency and fast response; suitable for VLA model inference):
  • Robotic arm joint motors
  • End-effector motor
• servo_control (servo mode; suitable for teleoperation):
  • High-frequency control loop (250Hz by default) with real-time smoothing
  • The move_end_pose interface can control the robotic arm and end effector simultaneously
• planning_control (planning mode; suitable for upper-layer trajectory planning):
  • Point-to-point planning: PTP planning algorithm and rrt_connect planning algorithm
  • Continuous waypoints planning
  • linear tool-end linear trajectory
  • circle tool-end circular trajectory
• mit_control (direct motor control using a hybrid torque-and-position control mode):
  • Robotic arm joint motors
  • End-effector motor