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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Reinforcement learning (RL) has demonstrated remarkable potential in roboticmanipulation but faces challenges in sample inefficiency and lack ofinterpretability, limiting its applicability in real world scenarios. Enablingthe agent to gain a deeper understanding and adapt more efficiently to diverseworking scenarios is crucial, and strategic knowledge utilization is a keyfactor in this process. This paper proposes a Knowledge Capture, Adaptation,and Composition (KCAC) framework to systematically integrate knowledge transferinto RL through cross-task curriculum learning. KCAC is evaluated using a twoblock stacking task in the CausalWorld benchmark, a complex roboticmanipulation environment. To our knowledge, existing RL approaches fail tosolve this task effectively, reflecting deficiencies in knowledge capture. Inthis work, we redesign the benchmark reward function by removing rigidconstraints and strict ordering, allowing the agent to maximize total rewardsconcurrently and enabling flexible task completion. Furthermore, we define twoself-designed sub-tasks and implement a structured cross-task curriculum tofacilitate efficient learning. As a result, our KCAC approach achieves a 40percent reduction in training time while improving task success rates by 10percent compared to traditional RL methods. Through extensive evaluation, weidentify key curriculum design parameters subtask selection, transition timing,and learning rate that optimize learning efficiency and provide conceptualguidance for curriculum based RL frameworks. This work offers valuable insightsinto curriculum design in RL and robotic learning.

Knowledge capture, adaptation and composition (KCAC): A framework for cross-task curriculum learning in robotic manipulation