Attentional Action Selection Using Reinforcement Learning

Nocera, Dario Di; Finzi, Alberto; Rossi, Silvia; Staffa, Mariacarla

doi:10.1007/978-3-642-33093-3_37

Cited by 6 publications

(16 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In Di Nocera et al (2012), a Q-learning algorithm is used to tune and adapt the frequencies of sensors sampling, while action selection is obtained as a side effect of this attentional regulation. In the following, we first recall the Q-learning algorithm and then we detail its application to the attentional shifting problem.…”

Section: Methodsmentioning

confidence: 99%

“…Analogously to Di Nocera et al (2012), in this paper we consider a Softmax method that selects the action to be executed through a Boltzmann distribution (Sutton and Barto, 1998):

P_{a} (a ∣ s, Q) = \frac{\exp^{\frac{Q (s, a)}{τ}}}{\sum_{b ϵ A false(s false)} \exp^{\frac{Q (s, b)}{τ}}}

…”

Section: Methodsmentioning

confidence: 99%

“…Analogously to Di Nocera et al (2012), in our model, for each behavior b we introduce an Action Space A b representing the set of possible periods P b = { p b 1 , …, p b k } for that behavior. That is, an action a b is a possible assignment of a clock period p b which regulates the sampling rate and the activation frequency of the associated behavior.…”

Section: Motivated Rl For Attentional Shiftingmentioning

confidence: 99%

“…In order to represent the curiosity state into the state space, we reformulate the State Space S b of Di Nocera et al (2012) introducing a new parameter representing the degree of curiosity of the agent. In the extended framework, the state s b is determined by a triple ( c b , p b , σ b ), where, c b represents the level of curiosity of the system, p b is for the current clock period, and σ b is the current perceptive state of a behavior b .…”

Section: Motivated Rl For Attentional Shiftingmentioning

confidence: 99%

See 3 more Smart Citations

The role of intrinsic motivations in attention allocation and shifting

et al. 2014

Self Cite

View full text Add to dashboard Cite

The concepts of attention and intrinsic motivations are of great interest within adaptive robotic systems, and can be exploited in order to guide, activate, and coordinate multiple concurrent behaviors. Attention allocation strategies represent key capabilities of human beings, which are strictly connected with action selection and execution mechanisms, while intrinsic motivations directly affect the allocation of attentional resources. In this paper we propose a model of Reinforcement Learning (RL), where both these capabilities are involved. RL is deployed to learn how to allocate attentional resources in a behavior-based robotic system, while action selection is obtained as a side effect of the resulting motivated attentional behaviors. Moreover, the influence of intrinsic motivations in attention orientation is obtained by introducing rewards associated with curiosity drives. In this way, the learning process is affected not only by goal-specific rewards, but also by intrinsic motivations.

show abstract

Section: Methodsmentioning

confidence: 99%

“…Analogously to Di Nocera et al (2012), in this paper we consider a Softmax method that selects the action to be executed through a Boltzmann distribution (Sutton and Barto, 1998):

P_{a} (a ∣ s, Q) = \frac{\exp^{\frac{Q (s, a)}{τ}}}{\sum_{b ϵ A false(s false)} \exp^{\frac{Q (s, b)}{τ}}}

…”

Section: Methodsmentioning

confidence: 99%

Section: Motivated Rl For Attentional Shiftingmentioning

confidence: 99%

Section: Motivated Rl For Attentional Shiftingmentioning

confidence: 99%

See 2 more Smart Citations

The role of intrinsic motivations in attention allocation and shifting

et al. 2014

Self Cite

View full text Add to dashboard Cite

show abstract

“…Typically, within these approaches, RL is used to directly model and optimize action selection strategies. Differently from these approaches, starting from a previous work [Di Nocera et al, 2012], we propose a model of action selection, which is indirectly provided by the learning of attentional allocation strategies by means of intrinsically motivated RL, where the states exploration policy is modulated by the curiosity. We refer to the model of attentional mechanisms proposed by [Burattini et al, 2010], while for the curiosity model we take inspiration from the model considered by [Litman, 2005].…”

Section: Introductionmentioning

confidence: 99%

Attentional Shifting and Curiosity: a Reinforcement Learning Approach

Nocera¹,

Finzi²,

Rossi³

et al. 2016

Anais Do 11. Congresso Brasileiro De Inteligência Computacional

Self Cite

View full text Add to dashboard Cite

Attention allocation strategies represent key capabilities of human beings, which are strictly connected with action selection and execution mechanisms, while intrinsic motivations directly affect the allocation of attentional resources. In this paper we propose a model of Reinforcement Learning (RL), where both these capabilities are involved. RL is deployed to learn how to allocate attentional resources in a behavior-based robotic system, while action selection is obtained as a side effect of the resulting motivated attentional behaviors. Moreover, the influence of intrinsic motivations in attention orientation is obtained by introducing internal rewards associated with curiosity drives. In this way, the learning process is affected not only by goalspecific rewards, but also by intrinsic motivations depending on the internal state of the system.

show abstract

Human Inspiration and Comparison for Monitoring Strategies in a Robotic Convoy Task

Rossi

Staffa

2014

From Animals to Animats 13

View full text Add to dashboard Cite

Attentional Action Selection Using Reinforcement Learning

Cited by 6 publications

References 6 publications

The role of intrinsic motivations in attention allocation and shifting

The role of intrinsic motivations in attention allocation and shifting

Attentional Shifting and Curiosity: a Reinforcement Learning Approach

Human Inspiration and Comparison for Monitoring Strategies in a Robotic Convoy Task

Contact Info

Product

Resources

About