Spot improves humanoid robot control with enhanced spatial awareness
SPOT: Spatial Perception-Oriented Long-Horizon Humanoid Teleoperation
Robotics
Summary
Controlling humanoid robots remotely is hard because operators often lose sight of important details like where objects or the robot itself are. The authors present SPOT, a system that gives operators a wider, more stable view that doesn’t interfere with robot movement. It uses special cameras and virtual reality technology to let users look around freely while controlling the robot. Tests show SPOT helps users complete tasks faster and more accurately, making it easier to teach robots how to do things over a long time.
What this means in practice
- •For robot operators: Control humanoid robots more effectively during complex tasks by maintaining a stable, wide view that enhances spatial awareness without interfering with robot motion.
- •For robot training teams: Collect higher-quality demonstration data for teaching humanoid robots by enabling operators to perceive and react to peripheral and task-relevant spatial information over long periods.
Authors
Lixing Fang, Ziyan Xiong, Sunli Chen, Zhiyang Dou, Chuang Gan
Abstract
High-quality demonstration data is becoming a central bottleneck for training general-purpose humanoid robots. While recent humanoid teleoperation systems have made substantial progress in retargeting human motion to robot motion, long-horizon loco-manipulation requires another capability: operators must maintain task-relevant spatial awareness over time, e.g., object locations, surrounding environments, the robot's pose. We call the extent of this awareness the operator's perceptual horizon. However, existing methods often shorten this: narrow views miss peripheral events, robot-mounted cameras become unstable during locomotion, and coupled head-view control makes looking around interfere with robot motion. We present SPOT, a Spatial Perception-Oriented VR Teleoperation system for collecting long-horizon humanoid demonstration data by providing extended perceptual horizon. SPOT combines a robot-mounted binocular fisheye camera, a wide-field stereoscopic display, viewpoint-decoupled free-looking, and visual stabilization to provide a robot-centric view that is wide, stable, and actively inspectable. Unlike conventional egocentric interfaces, SPOT decouples visual exploration from robot actuation: the egocentric stereo observation is rendered on a virtual hemisphere around the operator, so natural head rotations change where the operator looks within the wide-field view rather than commanding the robot head, camera, or torso. We evaluate SPOT on perception-critical humanoid data-collection tasks spanning drop recovery, peripheral retrieval, large-workspace bimanual manipulation, fine alignment, and dynamic interaction. SPOT improves efficiency, accuracy, and recovery speed, demonstrating its effectiveness for user-friendly and scalable long-horizon humanoid data collection.