Operating environments: ASD retail stores in China, WholeBodyVLA physical research evaluation workspace
Control modes: Autonomous
Terrain bag handover - Research demonstration - 2025-12-15 - Receive a bag, follow a person over steps, foam, wooden planks, gravel and artificial turf, then return the bag. - environment: WholeBodyVLA physical research evaluation workspace - Hardware: X2 prototype with 7-DoF arms and Omnipicker grippers; Head-mounted Intel RealSense D435i RGB-D camera - Software: WholeBodyVLA policy and execution decoder; LMO reinforcement-learning lower-body controller - Control: autonomous - source: arxiv.org
Arrow marker route - Research demonstration - 2025-12-15 - Follow a floor trajectory marked by visual arrows. - environment: WholeBodyVLA physical research evaluation workspace - Hardware: X2 prototype with 7-DoF arms and Omnipicker grippers; Head-mounted Intel RealSense D435i RGB-D camera - Software: WholeBodyVLA policy and execution decoder; LMO reinforcement-learning lower-body controller - Control: autonomous - source: arxiv.org
Cart pushing - Research demonstration - 2025-12-15 - Grasp a cart handle and push the loaded cart forward several metres. - environment: WholeBodyVLA physical research evaluation workspace - Hardware: X2 prototype with 7-DoF arms and Omnipicker grippers; Head-mounted Intel RealSense D435i RGB-D camera - Software: WholeBodyVLA policy and execution decoder; LMO reinforcement-learning lower-body controller - Control: autonomous - source: arxiv.org
Snacks to drawer - Research demonstration - 2025-12-15 - Walk to a table, grasp two snack bags, carry them to a drawer, place them inside and close the drawer. - environment: WholeBodyVLA physical research evaluation workspace - Hardware: X2 prototype with 7-DoF arms and Omnipicker grippers; Head-mounted Intel RealSense D435i RGB-D camera - Software: WholeBodyVLA policy and execution decoder; LMO reinforcement-learning lower-body controller - Control: autonomous - source: arxiv.org
Natural-language interaction - Customer-site use - 2026-09-28 - Greet shoppers, ask about their cooking needs, explain cookware and recommend products before handing transaction questions to store staff. - environment: ASD retail stores in China - Control: unspecified - Conditions: Human sales staff handle promotions, inventory and transactions - source: agibot.com
Bag packing - Research demonstration - 2025-12-15 - Grasp a paper bag with both hands, sidestep to a carton, squat and place the bag inside. - environment: WholeBodyVLA physical research evaluation workspace - Hardware: X2 prototype with 7-DoF arms and Omnipicker grippers; Head-mounted Intel RealSense D435i RGB-D camera - Software: WholeBodyVLA policy and execution decoder; LMO reinforcement-learning lower-body controller - Control: autonomous - source: arxiv.org
Box loading - Research demonstration - 2025-12-15 - Squat to grasp a box, stand, turn and place the box onto a cart. - environment: WholeBodyVLA physical research evaluation workspace - Hardware: X2 prototype with 7-DoF arms and Omnipicker grippers; Head-mounted Intel RealSense D435i RGB-D camera - Software: WholeBodyVLA policy and execution decoder; LMO reinforcement-learning lower-body controller - Control: autonomous - source: arxiv.org
Vacuum paper - Research demonstration - 2025-12-15 - Use a vacuum cleaner to remove paper debris from the floor. - environment: WholeBodyVLA physical research evaluation workspace - Hardware: X2 prototype with 7-DoF arms and Omnipicker grippers; Head-mounted Intel RealSense D435i RGB-D camera - Software: WholeBodyVLA policy and execution decoder; LMO reinforcement-learning lower-body controller - Control: autonomous - source: arxiv.org
Wipe coffee - Research demonstration - 2025-12-15 - Wipe coffee stains from a tabletop with a cloth. - environment: WholeBodyVLA physical research evaluation workspace - Hardware: X2 prototype with 7-DoF arms and Omnipicker grippers; Head-mounted Intel RealSense D435i RGB-D camera - Software: WholeBodyVLA policy and execution decoder; LMO reinforcement-learning lower-body controller - Control: autonomous - source: arxiv.org