2021
DOI: 10.1007/978-3-030-87361-5_15
|View full text |Cite
|
Sign up to set email alerts
|

Hindsight Curriculum Generation Based Multi-Goal Experience Replay

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
5
0

Year Published

2023
2023
2023
2023

Publication Types

Select...
2
1

Relationship

0
3

Authors

Journals

citations
Cited by 3 publications
(5 citation statements)
references
References 6 publications
0
5
0
Order By: Relevance
“…Robotic manipulation, as a kind of challenging tasks, has been widely used to examine the performance of many different RL approaches, such as those based on experience relabeling [2], [36], intrinsic motivation [35], [6], and guided exploration and exploitation [9], [23]. The basic idea behind these approaches is to improve the exploration efficiency, which is crucial in sparse-reward RL settings, where effective exploration is extremely difficult in manipulation tasks due to the sparsity of the goal space and the uninformative sparse rewards.…”
Section: A Manipulation With Sparse-reward Rlmentioning
confidence: 99%
See 2 more Smart Citations
“…Robotic manipulation, as a kind of challenging tasks, has been widely used to examine the performance of many different RL approaches, such as those based on experience relabeling [2], [36], intrinsic motivation [35], [6], and guided exploration and exploitation [9], [23]. The basic idea behind these approaches is to improve the exploration efficiency, which is crucial in sparse-reward RL settings, where effective exploration is extremely difficult in manipulation tasks due to the sparsity of the goal space and the uninformative sparse rewards.…”
Section: A Manipulation With Sparse-reward Rlmentioning
confidence: 99%
“…Guided exploration extends the experience relabeling approaches by creating an implicit curriculum of hindsight goals to lead the exploration towards target goals. The guidance metric can be calculated using Euclidean distance [23], [9] or other customized distances [3].…”
Section: A Manipulation With Sparse-reward Rlmentioning
confidence: 99%
See 1 more Smart Citation
“…DRL has attracted a lot of attention from diverse realworld applications including games [54], [55], [56], [57], robotics [58], [59], [60], [61], transportation [19], [20], construction [62], healthcare [63], navigation [22], [64], and etc [65]. These studies are grounded in domain knowledge and use DRL as a method to solve optimization and control problems under complex and dynamic environments.…”
Section: Drl Applicationsmentioning
confidence: 99%
“…Although reinforcement learning (RL) has demonstrated excellent performance on learning a single task, e.g., playing Go (Silver et al, 2016), robotic control (Schulman Figure 1: Main steps and components of CoTASP. et al, 2017;Degrave et al, 2022), and offline policy optimization (Yu et al, 2020;Yang et al, 2022), it still suffers from catastrophic forgetting and cross-task interference when learning an arbitrary sequence of tasks (McCloskey & Cohen, 1989;Bengio et al, 2020) or a curated sequence of tasks for a curriculum (Fang et al, 2019;Ao et al, 2021;. So it is challenging to train a meta-policy that can generalize to all learned tasks or even unseen ones with fast adaptation, which however is an inherent skill of human learning.…”
Section: Introductionmentioning
confidence: 99%