Telexistence and Physical Intelligence Announce Partnership to Automate Drink-Restocking Operations in Retail Stores
Telexistence Inc. (“TX”) and Physical Intelligence (“PI”) announced a partnership to develop robotic foundation models, aiming to automate tasks of drink-restocking operations in convenience stores.
To date, TX’s AI robot “TX GHOST” has been running in convenience stores to automate the majority of drink-restocking tasks. Here, teleoperation by remote operators was also combined, due to some error cases being unpredictable and having numerous patterns of its recovery process. For example, recovering rolled over beverages inside display shelves is one of the error cases that was challenging to automate with off-the-shelf technology.
By combining TX’s large amount of unique “embodied” teleoperation data and know-how with PI’s state-of-the-art general-purpose robotics foundation models, this partnership intends to shift these manual teleoperation tasks to fully autonomous operations.
Partnership scope:
Telexistence (TX) will provide robots running in retail stores along with its teleoperation data, and work on combining PI’s vision language action (VLA) models using TX’s proprietary robotics technology.
Physical Intelligence (PI) will develop policies that address the challenges of recovering from error cases by training PI’s VLA models on data generated during real production.
Going further, by establishing and running a training loop of injecting new operation data, training, and re-deploying VLA models, TX and PI aim at creating a platform where robots with human-like “physical intelligence” can react and autonomously perform physical labor at various industries.
Our partnership is already underway, and we will share further details as the work progresses.
About Physical Intelligence ( https://www.physicalintelligence.company/ )
Physical Intelligence is bringing general-purpose AI into the physical world. We are a group of engineers, scientists, roboticists, and company builders developing foundation models and learning algorithms to power the robots of today and the physically-actuated devices of the future.
Address: 396 Treat Avenue, San Francisco
Representative: Karol Hausman, Co-Founder & CEO
Founded: February 21, 2024
