Autonomous and fast robot learning through motivation

Rodríguez, Miguel Ángel; Iglesias, Roberto; Regueiro, Carlos V.; Correa, J. L. del Río; Barro, Senén

doi:10.1016/j.robot.2007.05.005

Cited by 23 publications

(13 citation statements)

References 2 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Mobile robots provide an essential tool when investigating the interaction of cognitive architectures and the physical environment. Robots have been used to investigate many different aspects of artificial intelligence such as mapping and localization techniques [15,16] , robot perception [17] and robot learning [18,19] . The reported research builds on the previous use of a mobile robot to investigate a specific area of cognitive science known as the anchoring problem [20] .…”

Section: Cognitive Architectures and Robotsmentioning

confidence: 99%

Probabilistic BDI in a Cognitive Robot Architecture

Davis

Miri

2012

IJCSAI

View full text Add to dashboard Cite

Abstract-The probabilistic approach to cognition has become an established approach in recent decades. Cognition is better viewed as solving probabilistic, rather than logical, inference problems; i.e. cognition is better understood in terms of probability theory, rather than in terms of logic. This article presents a cognitive architecture used to govern a robot probabilistically. The design and implementation of cognitive architectures is a useful tool for understanding cognition in a situated agent. The Cerno research project extended the CAMAL (Computational Architectures for Motivation, Affect, and Learning) model, by incorporating probabilistic reasoning in its BDI model. Subsequent development of CAMAL has integrated all the valanced affective predicates across the architecture. Extensive experiments with synthetic and real robots demonstrate an improvement in the overall performance, success rate, task effectiveness, and goal achievement of the cognitive architecture.

show abstract

Section: Cognitive Architectures and Robotsmentioning

confidence: 99%

Probabilistic BDI in a Cognitive Robot Architecture

Davis

Miri

2012

IJCSAI

View full text Add to dashboard Cite

show abstract

“…Reinforcement Learning is suitable to get robots learning from their own experiences and environment interaction. Nevertheless, from our previous work [2], [3], [4], [5], [6] we know that reinforcement learning is too slow and requires too many trials to learn a particular task. This makes its application on a real robot almost impossible.…”

Section: Achieving Fast Learning Processesmentioning

confidence: 99%

Learning in real robots from environment interaction

Vidal¹,

Rodríguez²,

González³

et al. 2012

jopha

View full text Add to dashboard Cite

Abstract-This article describes a proposal to achieve fast robot learning from its interaction with the environment. Our proposal will be suitable for continuous learning procedures as it tries to limit the instability that appears every time the robot encounters a new situation it had not seen before. On the other hand, the user will not have to establish a degree of exploration (usual in reinforcement learning) and that would prevent continual learning procedures. Our proposal will use an ensemble of learners able to combine dynamic programming and reinforcement learning to predict when a robot will make a mistake. This information will be used to dynamically evolve a set of control policies that determine the robot actions.Index Terms-continuous robot learning, robot adaptation, learning from environment interaction, reinforcement learning.

show abstract

“…RL has attracted attention and been widely used in intelligent robot control systems especially in the field of autonomous mobile robots [6], [8]- [11]. Without knowing which actions to take, the reinforcement learning agent exploits and explores actions to discover which action gives the most reward in the long run.…”

Section: Introductionmentioning

confidence: 99%

The residual gradient FACL algorithm for differential games

Awheda

Schwartz

2015

2015 IEEE 28th Canadian Conference on Electrical and Computer Engineering (CCECE)

View full text Add to dashboard Cite

A new fuzzy reinforcement learning algorithm that tunes the input and the output parameters of a fuzzy logic controller is proposed in this paper. The proposed algorithm uses three fuzzy inference systems (FISs); one is used as an actor (fuzzy logic controller, FLC), and the other two FISs are used as critics. The proposed algorithm uses the residual gradient value iteration algorithm described in [4] to tune the input and the output parameters of the actor (FLC) of the learning robot. The proposed algorithm also tunes the input and the output parameters of the critics. The proposed algorithm is called the residual gradient fuzzy actor critics learning (RGFACL) algorithm. The proposed algorithm is used to learn a single pursuit-evasion differential game. Simulation results show that the performance of the proposed RGFACL algorithm outperforms the performance of the fuzzy actor critic learning (FACL) and the Q-learning fuzzy inference system (QLFIS) algorithms proposed in [3] and [7], respectively, in terms of convergence and speed of learning.

show abstract

Autonomous and fast robot learning through motivation

Cited by 23 publications

References 2 publications

Probabilistic BDI in a Cognitive Robot Architecture

Probabilistic BDI in a Cognitive Robot Architecture

Learning in real robots from environment interaction

The residual gradient FACL algorithm for differential games

Contact Info

Product

Resources

About