Real-world reinforcement learning for autonomous humanoid robot docking

Navarro-Guerrero, Nicolás; Weber, Cornelius; Schroeter, Pascal; Wermter, Stefan

doi:10.1016/j.robot.2012.05.019

Cited by 31 publications

(18 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Many artificial intelligent techniques have been proposed by scholars for mobile robot path planning, such as reinforcement learning, fuzzy logic, and genetic algorithm. Reinforcement learning algorithm [1,5,6] has simple and complete theory, but it is mostly used in static environment because its infinite state in complex environment. Fuzzy logic control (FLC) [7][8][9] has the capacity to handle uncertain and imprecise information obtained from sensors using linguistic rules.…”

Section: The Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Mobile Robot Path Planning Using Polyclonal-Based Artificial Immune Network

Deng

2013

Journal of Control Science and Engineering

View full text Add to dashboard Cite

Polyclonal based artificial immune network (PC-AIN) is utilized for mobile robot path planning. Artificial immune network (AIN) has been widely used in optimizing the navigation path with the strong searching ability and learning ability. However, artificial immune network exists as a problem of immature convergence which some or all individuals tend to the same extreme value in the solution space. Thus, polyclonal-based artificial immune network algorithm is proposed to solve the problem of immature convergence in complex unknown static environment. Immunity polyclonal algorithm (IPCA) increases the diversity of antibodies which tend to the same extreme value and finally selects the antibody with highest concentration. Meanwhile, immunity polyclonal algorithm effectively solves the problem of local minima caused by artificial potential field during the structure of parameter in artificial immune network. Extensive experiments show that the proposed method not only solves immature convergence problem of artificial immune network but also overcomes local minima problem of artificial potential field. So, mobile robot can avoid obstacles, escape traps, and reach the goal with optimum path and faster convergence speed.

show abstract

Section: The Related Workmentioning

confidence: 99%

“…Experimental environment sets are as follows: the initial position of robot (5,4), the goal position (10,16), the number of obstacles are 5, the obstacles position (9, 10), (7,6), (5,8), (8,13), and (6, 12), respectively. The velocity is 0.1 m/s.…”

Section: Performance Comparison Of Ain and Pc-ainmentioning

confidence: 99%

Mobile Robot Path Planning Using Polyclonal-Based Artificial Immune Network

Deng

2013

Journal of Control Science and Engineering

View full text Add to dashboard Cite

show abstract

“…In [18] supervised reinforcement learning is used for autonomous humanoid robot docking. It uses Gaussian distributed states activation so inputs can be continuous, however the action space is discrete and there are only 4 actions.…”

Section: Introductionmentioning

confidence: 99%

Supervised fuzzy reinforcement learning for robot navigation

Fathinezhad

Derhami

Rezaeian

2016

Applied Soft Computing

View full text Add to dashboard Cite

This paper addresses a new method for combination of supervised learning and Reinforcement Learning (RL). Applying supervised learning in robot navigation encounters serious challenges such as inconsistent and noisy data, difficulty for gathering training data, and high error in training data. RL capabilities such as training only by one evaluation scalar signal, and high degree of exploration have encouraged researchers to use RL in robot navigation problem. However, RL algorithms are time consuming as well as suffer from high failure rate in the training phase. Here, we propose Supervised Fuzzy Sarsa Learning (SFSL) as a novel idea for utilizing advantages of both supervised and reinforcement learning algorithms. A zero order Takagi-Sugeno fuzzy controller with some candidate actions for each rule is considered as the main module of robot"s controller. The aim of training is to find the best action for each fuzzy rule. In the first step, a human supervisor drives an Epuck robot within the environment and the training data are gathered. In the second step as a hard tuning, the training data are used for initializing the value (worth) of each candidate action in the fuzzy rules. Afterwards, the fuzzy Sarsa learning module, as a critic-only based fuzzy reinforcement learner, fine tunes the parameters of conclusion parts of the fuzzy controller online. The proposed algorithm is used for driving E-puck robot in the environment with obstacles. The experiment results show that the proposed approach decreases the learning time and the number of failures; also it improves the quality of the robot"s motion in the testing environments.

show abstract

“…Many autonomous charging systems have already been developed for a variety of mobile robotic systems, such as wheeled robots [1][2][3][4][5][6], humanoids [7], unmanned aerial vehicles [8], cars [9], and autonomous underwater vehicles (AUVs) [10][11][12]. These robotic systems have the potential to greatly aid the implementation of critical services, such as surveillance [3,4], home sanitation [1], and environmental monitoring [10], without the need for human intervention.…”

Section: Introductionmentioning

confidence: 99%