Online Gait Learning for Modular Robots with Arbitrary Shapes and Sizes

Weel, Berend; d’Angelo, Massimiliano; Haasdijk, Evert; Eiben, A. E.

doi:10.1162/artl_a_00223

Cited by 8 publications

(9 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Most of the mentioned research considers the off-line development of locomotive controllers, i.e., controller optimization as a separate phase before deployment intending to developing controllers that remain fixed once deployed. Weel et al (2017) considered on-line gait learning, where the controller is adapted to the robot's task environment during deployment. The experiments showed that spline-based controllers with the RL PoWER algorithm provide dynamic autonomous on-line gait learning capabilities.…”

Section: Lifetime Locomotion Learningmentioning

confidence: 99%

“…Therefore, we adopt a system architecture that does contain such a phase. This architecture, called the Triangle of Life, was introduced in Eiben et al (2013) and used in several experimental studies that addressed the task of gait learning (Rossi and Eiben, 2014;Jelisavcic et al, 2016;Weel et al, 2017). Note that the choice of the gait learning task is not arbitrary.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Lamarckian Evolution of Simulated Modular Robots

et al. 2019

Self Cite

View full text Add to dashboard Cite

We study evolutionary robot systems where not only the robot brains but also the robot bodies are evolvable. Such systems need to include a learning period right after 'birth' to acquire a controller that fits the newly created body. In this paper we investigate the possibility of bootstrapping infant robot learning through employing Lamarckian inheritance of parental controllers. In our system controllers are encoded by a combination of a morphology dependent component, a Central Pattern Generator (CPG), and a morphology independent part, a Compositional Pattern Producing Network (CPPN). This makes it possible to transfer the CPPN part of controllers between different morphologies and to create a Lamarckian system. We conduct experiments with simulated modular robots whose fitness is determined by the speed of locomotion, establish the benefits of inheriting optimized parental controllers, shed light on the conditions that influence these benefits, and observe that changing the way controllers are evolved also impacts the evolved morphologies.

show abstract

Section: Lifetime Locomotion Learningmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Lamarckian Evolution of Simulated Modular Robots

et al. 2019

Self Cite

View full text Add to dashboard Cite

show abstract

“…], [6]. The properties of this algorithm were investigated in comparison to HyperNEAT and Simulated Annealing [7], [19]. These studies have shown that RL PoWER is a superior method for on-line gait learning since it converges quickly to learn sufficiently good gaits in a short time.…”

Section: Related Workmentioning

confidence: 99%

“…Technically speaking, we are interested in a general gait learning algorithm that works for any given robot within the space of all possible morphologies constructible with the modules we use. Our algorithm of choice is the RL PoWER algorithm as proposed by Kober and Peters [6] and employed for gait learning in Roombots in our previous work [7].…”

Section: Introductionmentioning

confidence: 99%

Improving RL power for on-line evolution of gaits in modular robots

Jelisavcic

Carlo

Haasdijk

et al. 2016

2016 IEEE Symposium Series on Computational Intelligence (SSCI)

Self Cite

View full text Add to dashboard Cite

This paper addresses the problem of on-line gait learning in modular robots whose shape is not known in advance. The best algorithm for this problem known to us is a reinforcement learning method, called RL PoWER. In this study we revisit the original RL PoWER algorithm and observe that in essence it is a specific evolutionary algorithm. Based on this insight we propose two modifications of the main search operators and compare the quality of the evolved gaits when either or both of these modified operators are employed. The results show that using 2-parent crossover as well as mutation with selfadaptive step-sizes can significantly improve the performance of the original algorithm.

show abstract

“…23 Because of low cost, time consumption, and high precision of online learning for large-scale samples, online learning is applied to multiple fields, such as dynamic two categories, 24 vowel imitation learning, 25 and modular robots. 26 In this article, the smooth iterative online support tensor machine (SIOSTM) algorithm was developed by combining the two algorithms of linear SHTM and online stochastic gradient descent (OSGD). Using the tensor-type data as input, the characteristic parameters of the fault samples of the electric vehicle extension were extracted and the fault parameters were diagnosed.…”

Section: Introductionmentioning

confidence: 99%

Smooth iteration online support tension machine algorithm and application in fault diagnosis of electric vehicle extended range

Zhang

et al. 2018

Advances in Mechanical Engineering

View full text Add to dashboard Cite

It is difficult to develop accurate mathematical models to describe the range extender electric vehicles due to the nonlinear and complex coupling of the monitoring signal sources resulted from the massive moving parts and complex architecture in range extender and the limited storage space of the diagnostic device. In this study, we proposed the smooth iterative online support tensor machine algorithm, which is combined with support higher-order tensor machine and online stochastic gradient descent method, and applied it to the fault diagnosis of the range indicator. Four methods with different algorithms, support vector machine, smooth iterative online support vector machine, linear support higherorder tensor machine, and smooth iterative online support tensor machine algorithms, were adopted to diagnose and classify the fault samples of the range indicator by comparing the diagnostic accuracy and model learning time. It is found that the fault diagnosis method based on the smooth iterative online tensor machine showed a higher accuracy, shorter learning time, and less storage space. Based on the experimental results, it is feasible to apply smooth iterative online support tensor machine model to the fault diagnosis of electric vehicle extenders.

show abstract

Online Gait Learning for Modular Robots with Arbitrary Shapes and Sizes

Cited by 8 publications

References 25 publications

Lamarckian Evolution of Simulated Modular Robots

Lamarckian Evolution of Simulated Modular Robots

Improving RL power for on-line evolution of gaits in modular robots

Smooth iteration online support tension machine algorithm and application in fault diagnosis of electric vehicle extended range

Contact Info

Product

Resources

About