English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

We propose a novel general method that finds action-grounded, discrete objectand effect categories and builds probabilistic rules over them for non-trivialaction planning. Our robot interacts with objects using an initial actionrepertoire that is assumed to be acquired earlier and observes the effects itcan create in the environment. To form action-grounded object, effect, andrelational categories, we employ a binary bottleneck layer in a predictive,deep encoder-decoder network that takes the image of the scene and the actionapplied as input, and generates the resulting effects in the scene in pixelcoordinates. After learning, the binary latent vector represents action-drivenobject categories based on the interaction experience of the robot. To distillthe knowledge represented by the neural network into rules useful for symbolicreasoning, a decision tree is trained to reproduce its decoder function.Probabilistic rules are extracted from the decision paths of the tree and arerepresented in the Probabilistic Planning Domain Definition Language (PPDDL),allowing off-the-shelf planners to operate on the knowledge extracted from thesensorimotor experience of the robot. The deployment of the proposed approachfor a simulated robotic manipulator enabled the discovery of discreterepresentations of object properties such as `rollable' and `insertable'. Inturn, the use of these representations as symbols allowed the generation ofeffective plans for achieving goals, such as building towers of the desiredheight, demonstrating the effectiveness of the approach for multi-step objectmanipulation. Finally, we demonstrate that the system is not only restricted tothe robotics domain by assessing its applicability to the MNIST 8-puzzle domainin which learned symbols allow for the generation of plans that move the emptytile into any given position.

DeepSym: Deep Symbol Generation and Rule Learning from Unsupervised Continuous Robot Interaction for Planning