{"id":880,"date":"2023-11-02T11:34:10","date_gmt":"2023-11-02T11:34:10","guid":{"rendered":"https:\/\/todaysainews.com\/index.php\/2023\/11\/02\/stacking-our-way-to-more-general-robots-2\/"},"modified":"2025-04-27T07:30:58","modified_gmt":"2025-04-27T07:30:58","slug":"stacking-our-way-to-more-general-robots-2","status":"publish","type":"post","link":"https:\/\/todaysainews.com\/index.php\/2023\/11\/02\/stacking-our-way-to-more-general-robots-2\/","title":{"rendered":"Stacking our way to more general robots"},"content":{"rendered":"

[ad_1]
\n<\/p>\n

\n
\n
\n

Research<\/p>\n

\n
Published<\/dt>\n
\n
Authors<\/dt>\n
\n

The Robotics Team<\/p>\n<\/dd>\n<\/dl>\n

\n<\/section><\/div>\n\"A\n <\/picture>\n<\/p><\/div>\n
\n

Introducing RGB-Stacking as a new benchmark for vision-based robotic manipulation<\/b><\/p>\n

Picking up a stick and balancing it atop a log or stacking a pebble on a stone may seem like simple \u2014 and quite similar \u2014 actions for a person. However, most robots struggle with handling more than one such task at a time. Manipulating a stick requires a different set of behaviours than stacking stones, never mind piling various dishes on top of one another or assembling furniture. Before we can teach robots how to perform these kinds of tasks, they first need to learn how to interact with a far greater range of objects. As part of DeepMind\u2019s mission<\/a> and as a step toward making more generalisable and useful robots, we\u2019re exploring how to enable robots to better understand the interactions of objects with diverse geometries.<\/p>\n<\/div>\n

\n<\/figure>\n
\n

In a paper to be presented at CoRL 2021<\/a> (Conference on Robot Learning) and available now as a preprint on OpenReview<\/a>, we introduce RGB-Stacking as a new benchmark for vision-based robotic manipulation. In this benchmark, a robot has to learn how to grasp different objects and balance them on top of one another. What sets our research apart from prior work is the diversity of objects used and the large number of empirical evaluations performed to validate our findings. Our results demonstrate that a combination of simulation and real-world data can be used to learn complex multi-object manipulation and suggest a strong baseline for the open problem of generalising to novel objects. To support other researchers, we\u2019re open-sourcing<\/a> a version of our simulated environment, and releasing the designs<\/a> for building our real-robot RGB-stacking environment, along with the RGB-object models and information for 3D printing them. We are also open-sourcing a collection of libraries and tools<\/a> used in our robotics research more broadly.<\/p>\n<\/div>\n

\n

RGB-Stacking benchmark<\/p>\n<\/figcaption><\/figure>\n

\n

With RGB-Stacking, our goal is to train a robotic arm via reinforcement learning to stack objects of different shapes. We place a parallel gripper attached to a robot arm above a basket, and three objects in the basket \u2014 one red, one green, and one blue, hence the name RGB. The task is simple: stack the red object on top of the blue object within 20 seconds, while the green object serves as an obstacle and distraction. The learning process ensures that the agent acquires generalised skills through training on multiple object sets. We intentionally vary the grasp and stack affordances \u2014 the qualities that define how the agent can grasp and stack each object. This design principle forces the agent to exhibit behaviours that go beyond a simple pick-and-place strategy.<\/p>\n<\/div>\n

\n

Each triplet poses its own unique challenges to the agent: Triplet 1 requires a precise grasp of the top object; Triplet 2 often requires the top object to be used as a tool to flip the bottom object before stacking; Triplet 3 requires balancing; Triplet 4 requires precision stacking (i.e., the object centroids need to align); and the top object of Triplet 5 can easily roll off if not stacked gently. In assessing the challenges of this task, we found that our hand-coded scripted baseline had a 51% success rate at stacking.<\/p>\n<\/figcaption><\/figure>\n

\n

Our RGB-Stacking benchmark includes two task versions with different levels of difficulty. In \u201cSkill Mastery,\u201d our goal is to train a single agent that\u2019s skilled in stacking a predefined set of five triplets. In \u201cSkill Generalisation,\u201d we use the same triplets for evaluation, but train the agent on a large set of training objects \u2014 totalling more than a million possible triplets. To test for generalisation, these training objects exclude the family of objects from which the test triplets were chosen. In both versions, we decouple our learning pipeline into three stages:<\/p>\n