Quadruped controller coordinates arm and base from end-effector trajectories

Quadruped controller coordinates arm and base from end-effector trajectories

Live•Nov 15, 12:30 AM•2 min read•
Carlos Mendez
Carlos Mendez

PolyMorphWBC coordinates a quadruped’s arm and base from end-effector trajectories for extended-reach manipulation across four families of 6-DoF mounted arms, first reported by Arxiv. The controller must decide how the base and arm share motion, with that choice depending on mounted-arm morphology; it can execute short-horizon manipulation trajectories beyond the initial fixed-base reach.

The task-level input is a short-horizon end-effector (EE) trajectory that specifies the desired manipulation motion without prescribing how the arm and mobile base realize it. A trajectory-conditioned coordinator uses previewed EE targets to determine base motion; arm-specific inverse kinematics (IK) tracks the current target, and a frozen locomotion backend executes the resulting motion.

PolyMorphWBC uses morphology conditioning derived from the Unified Robot Description Format (URDF), together with online residual-wrench feedback, to support cross-arm reuse. It was trained on procedurally generated arms and structured trajectories both within and beyond fixed-base reach.

In IsaacLab, it achieved 93.95% completion on held-out trajectories and arm morphologies, with 2.30cm position error and 4.81° orientation error. In MuJoCo evaluation against UMI-on-Legs, it reduced end-effector position error by 69–72% and orientation error by 91–97% across four within- and extended-reach trajectories.

The results pair trajectory-error measurements with a physical pick-and-place demonstration. On a physical Go2–ARX5 system, PolyMorphWBC demonstrated language-conditioned extended-reach pick-and-place using EE trajectories from pi_0, without vision-language-action (VLA)-generated base commands.

OpenAI’s ChatGPT Wallet lets Dots agents shop with approval

OpenAI is rolling out ChatGPT Wallet, letting its Dots agents make purchases with a user’s card, first reported by The Verge. The initial rollout is limited to US ChatGPT Pro subscribers aged 18 and older with a US billing address, and every checkout still requires user approval and a security check.

Eligible US-issued American Express, Chase, Mastercard and Visa cards can be added. Once a card is in the wallet, a Dots agent can make a purchase in response to a prompt, but it must check in before every checkout. OpenAI’s support page says each purchase also requires biometric verification, such as Face ID or a fingerprint scan.

After the user reviews and approves a purchase, ChatGPT Wallet substitutes a payment token for the credit card details. OpenAI says the actual card information is not shared with the Dots agent or the business receiving the payment, in a process similar to tap to pay, Apple Pay or Google Wallet.

For now, purchases remain user-approved; OpenAI eventually expects users to delegate them under preset rules. Those rules could include spending limits, purchase categories or time periods, with users pre-approving purchases within those parameters.

The wallet may not work everywhere: some retailers restrict or block AI agents, and Amazon blocked Meta’s Muse agents earlier this year. OpenAI says websites need to ensure agents are not inadvertently blocked, while payment and risk systems need to recognize legitimate purchases made by agents.