Unitree has unveiled the AS2-W, a wheel-legged quadruped robot priced at approximately $36,700 — roughly half the base cost of Boston Dynamics' Spot — that uses real-time reinforcement learning to navigate cliffs, streams, and rough terrain1,2.
The company released demo footage this week showing the AS2-W crossing steep cliff faces and rocky streams, transitioning mid-stride from wheeled locomotion on flat ground to full quadruped stepping on broken terrain. The robot processes its environment in real time through reinforcement learning rather than pre-programmed path planning.
**Hardware specifications.** The AS2-W weighs 25 kg and is powered by 16 low-inertia motors. It combines articulated legs with wheels at each joint, enabling it to climb slopes up to 45°, clear 80 cm obstacles, and reach speeds of up to 6 m/s. Its payload capacity is 150 kg — a figure that positions it for industrial use cases where heavy sensor packages or tools must be carried across uneven ground.
The AS2-W is built on Unitree's AS2 quadruped base, which the company unveiled in February 2026. The wheel-legged variant adds the hybrid locomotion system.
**Pricing gap with Spot.** Boston Dynamics' Spot ranges from $75,000 to $375,000. At approximately $36,700, the AS2-W sits well below even Spot's base configuration. ANALYSIS That price differential could open the industrial quadruped market to smaller operators and startups that have been unable to justify Spot-tier capital expenditure.
Unitree describes the AS2-W's terrain-handling capability as driven by real-time reinforcement learning — an AI approach where the robot learns locomotion policies through trial and reward signals rather than explicit programming. The demo footage shows the system adapting to novel terrain features without human intervention.
ANALYSIS The AS2-W represents a direct challenge to Boston Dynamics' position in the industrial quadruped segment, combining a substantially lower price point with payload capacity and terrain-handling specifications that overlap with Spot's core use cases. The reinforcement-learning locomotion system is central to Unitree's pitch: the robot's ability to handle unprogrammed terrain in real time is the capability that makes the hardware viable for field deployment without dedicated engineering support on-site.