Object-Oriented Reinforcement Learning in Cooperative Multiagent Domains

Silva, Felipe Leno da; Glatt, Ruben; Costa, Anna Helena Reali

doi:10.1109/bracis.2016.015

Cited by 12 publications

(12 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A Best Paper Award from the BRACIS conference was awarded to one of those publications [Silva et al 2016]. Our work also received a Honorable Mention as Best Student Poster at the AAAI Conference on Artificial Intelligence in 2017.…”

Section: Scientific Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Methods and Algorithms for Knowledge Reuse in Multiagent Reinforcement Learning

Silva¹,

Costa²

2020

Anais Do Concurso De Teses E Dissertações Da SBC (CTD-SBC 2020)

Self Cite

View full text Add to dashboard Cite

Reinforcement Learning (RL) is a powerful tool that has been used to solve increasingly complex tasks. RL operates through repeated interactions of the learning agent with the environment, via trial and error. However, this learning process is extremely slow, requiring many interactions. In this thesis, we leverage previous knowledge so as to accelerate learning in multiagent RL problems. We propose knowledge reuse both from previous tasks and from other agents. Several flexible methods are introduced so that each of these two types of knowledge reuse is possible. This thesis adds important steps towards more flexible and broadly applicable multiagent transfer learning methods.

show abstract

Section: Scientific Resultsmentioning

confidence: 99%

“…The following full papers are the main related publications: [Silva et al 2020a, Silva et al 2020b, Silva and Costa 2019, Silva et al 2019b, Silva et al 2020c, Silva et al 2019a, Silva et al 2016.…”

Section: Scientific Resultsmentioning

confidence: 99%

Methods and Algorithms for Knowledge Reuse in Multiagent Reinforcement Learning

Silva¹,

Costa²

2020

Anais Do Concurso De Teses E Dissertações Da SBC (CTD-SBC 2020)

Self Cite

View full text Add to dashboard Cite

show abstract

“…While a common procedure in the literature is to provide hand-coded task mappings, autonomously computing task similarities is hard. In our BRACIS paper, we propose the Multiagent Object-Oriented MDP (MOO-MDP) (Silva, Glatt, and Costa 2016), a relational approach in which the state space is described through classes of objects. The main idea is that each entity in the environment (agent or not) belongs to a class of objects, and then the agent can generalize experiences.…”

Section: Partial Results and Future Workmentioning

confidence: 99%

Accelerating Multiagent Reinforcement Learning through Transfer Learning

Silva

Costa

2017

AAAI

View full text Add to dashboard Cite

Reinforcement Learning (RL) is a widely used solution for sequential decision-making problems and has been used in many complex domains. However, RL algorithms suffer from scalability issues, especially when multiple agents are acting in a shared environment. This research intends to accelerate learning in multiagent sequential decision-making tasks by reusing previous knowledge, both from past solutions and advising between agents. We intend to contribute a Transfer Learning framework focused on Multiagent RL, requiring as few domain-specific hand-coded parameters as possible.

show abstract

“…Our work seeks to present a solution to how agents might automatically generate expressive temporally extended auxiliary tasks--in contrast to intermediate tasks--that can maximally leverage the directed experience of a single-task curriculum in object-centric environments. TaskExplore is distinct from the limited task generation works in curriculum learning literature [16,17], where tasks are not manually designed. The goal of our approach is not to generate good intermediate tasks to obtain experience samples, as is the goal in task generation for curriculum learning [15].…”

Section: Related Workmentioning

confidence: 99%

Exploiting Contextual Structure to Generate Useful Auxiliary Tasks

Quartey¹,

Shah²,

Konidaris³

2023

Preprint

View full text Add to dashboard Cite

Reinforcement learning requires interaction with an environment, which is expensive for robots. This constraint necessitates approaches that work with limited environmental interaction by maximizing the reuse of previous experiences. We propose an approach that maximizes experience reuse while learning to solve a given task by generating and simultaneously learning useful auxiliary tasks. To generate these tasks, we construct an abstract temporal logic representation of the given task and leverage large language models to generate contextaware object embeddings that facilitate object replacements. Counterfactual reasoning and off-policy methods allow us to simultaneously learn these auxiliary tasks while solving the given target task. We combine these insights into a novel framework for multitask reinforcement learning and experimentally show that our generated auxiliary tasks share similar underlying exploration requirements as the given task, thereby maximizing the utility of directed exploration. Our approach allows agents to automatically learn additional useful policies without extra environment interaction.

show abstract

Object-Oriented Reinforcement Learning in Cooperative Multiagent Domains

Cited by 12 publications

References 11 publications

Methods and Algorithms for Knowledge Reuse in Multiagent Reinforcement Learning

Methods and Algorithms for Knowledge Reuse in Multiagent Reinforcement Learning

Accelerating Multiagent Reinforcement Learning through Transfer Learning

Exploiting Contextual Structure to Generate Useful Auxiliary Tasks

Contact Info

Product

Resources

About