Chapter 5: Vision Features

Choose the correct model: Ranger supports face-follow and gesture recognition, with motion feedback only in wheeled-legged mode. Switch modes before starting. The quadruped gesture actions and ball-grasping sequence below apply only to supported arm-equipped quadruped models; Ranger and Rider2 have no robotic arm.

Vision features are located in the 「🎮 Demos」 menu, including Face Follow, Gesture Recognition, and Ball Catch. The Demos menu has two pages: page 1 contains Face Follow, Gesture Recognition, etc.; page 2 contains Ball Catch, etc.

5.1 Face Follow

After detecting a face, the robot automatically rotates its body to keep the camera pointed at the detected face.

1. Enter 「🎮 Demos」 → 「Face Follow」

2. Stand 1–2 meters in front of the robot, facing the camera directly

3. The robot automatically detects the face position

4. Once a face is detected, the robot adjusts its body orientation to always face the person

5. Move left and right to observe the robot tracking your movement

6. Press C to exit back to the Demos menu

💡 Tips: Best results in well-lit environments, facing the camera directly, at a distance of 1–2 meters. Sunglasses or face masks may affect recognition.

5.2 Gesture Recognition

Make different hand gestures in front of the camera — the robot recognizes them and performs the corresponding action with sound effects.

Quadruped gesture actions (Ranger uses the separate mapping below):

Steps: 1. Enter 「🎮 Demos」 → 「Gesture Recognition」 2. Hold your hand in front of the camera, keeping the gesture steady for 2–3 seconds 3. The screen displays the recognized gesture icon 4. After confirmation (multiple consecutive frames recognized), the robot performs the action 5. The action name is displayed on screen during execution

💡 Tips: - Keep gestures steady; avoid rapid changes - Best detection at 30–50 cm from the camera - Each action has a cooldown period (minimum 5 seconds); changing gestures during cooldown won't trigger a new action - Cluttered backgrounds may reduce accuracy; a plain background is recommended

5.3 Ball Catch (supported arm-equipped quadrupeds only)

The robot automatically locates a colored ball, approaches it, and grabs it with its claw.

Steps: 1. Place a solid-colored ball (red, green, or blue) on the ground 30–50 cm in front of the robot 2. Enter 「🎮 Demos」 → 「Ball Catch」 3. Press A to cycle through target colors (Red → Green → Blue → Red) 4. Press B to adjust the grabbing distance 5. The currently selected color and distance are shown at the top of the screen 6. Once the settings are correct, press D to start 7. The robot automatically executes the following sequence: - Image recognition: locate the target-colored ball - Turn to align: rotate body to face the ball - Walk approach: move toward the ball using gait mode - Fine adjustment: precise posture adjustment at close range - Reach and grab: lower body and extend claw to grab the ball - Stand up: return to standing posture after grabbing

Key controls:

💡 Tips: - Use solid-colored balls (red, green, blue); multi-colored or patterned balls are harder to detect - Avoid direct strong light on the ball surface, as reflections may interfere with detection - Place the ball on flat ground with no same-colored objects nearby - Only models with a robotic arm support the grab function

Ranger gesture actions

Enter Demos → Gesture Recognition after switching Ranger to wheeled-legged mode. Hold the gesture steady for 2–3 seconds, around 30–50 cm from the camera. Recognition and sound feedback may appear without a movement in car mode. Face Follow also requires wheeled-legged mode for motion feedback.