## Distributed online learning of central pattern generators in modular robots

Publication: Research - peer-review › Article in proceedings – Annual report year: 2010

Without internal affiliation

### Standard

**Distributed online learning of central pattern generators in modular robots.** / Christensen, David Johan; Spröwitz, Alexander; Ijspeert, Auke Jan.

Publication: Research - peer-review › Article in proceedings – Annual report year: 2010

### Harvard

*From Animals to Animats 11.*Springer, pp. 402-412. Lecture Notes in Computer Science, no. 6226, , 10.1007/978-3-642-15193-4_38

### APA

*From Animals to Animats 11.*(pp. 402-412). Springer. (Lecture Notes in Computer Science; No. 6226). 10.1007/978-3-642-15193-4_38

### CBE

### MLA

*From Animals to Animats 11.*Springer. 2010. 402-412. (Lecture Notes in Computer Science; Journal number 6226). Available: 10.1007/978-3-642-15193-4_38

### Vancouver

### Author

### Bibtex

}

### RIS

TY - GEN

T1 - Distributed online learning of central pattern generators in modular robots

AU - Christensen,David Johan

AU - Spröwitz,Alexander

AU - Ijspeert,Auke Jan

PY - 2010

Y1 - 2010

N2 - In this paper we study distributed online learning of locomotion gaits for modular robots. The learning is based on a stochastic approximation method, SPSA, which optimizes the parameters of coupled oscillators used to generate periodic actuation patterns. The strategy is implemented in a distributed fashion, based on a globally shared reward signal, but otherwise utilizing local communication only. In a physics-based simulation of modular Roombots robots we experiment with online learning of gaits and study the effects of: module failures, different robot morphologies, and rough terrains. The experiments demonstrate fast online learning, typically 5-30 min. for convergence to high performing gaits (≈ 30 cm/sec), despite high numbers of open parameters (45-54). We conclude that the proposed approach is efficient, effective and a promising candidate for online learning on many other robotic platforms.

AB - In this paper we study distributed online learning of locomotion gaits for modular robots. The learning is based on a stochastic approximation method, SPSA, which optimizes the parameters of coupled oscillators used to generate periodic actuation patterns. The strategy is implemented in a distributed fashion, based on a globally shared reward signal, but otherwise utilizing local communication only. In a physics-based simulation of modular Roombots robots we experiment with online learning of gaits and study the effects of: module failures, different robot morphologies, and rough terrains. The experiments demonstrate fast online learning, typically 5-30 min. for convergence to high performing gaits (≈ 30 cm/sec), despite high numbers of open parameters (45-54). We conclude that the proposed approach is efficient, effective and a promising candidate for online learning on many other robotic platforms.

U2 - 10.1007/978-3-642-15193-4_38

DO - 10.1007/978-3-642-15193-4_38

M3 - Article in proceedings

SP - 402

EP - 412

BT - From Animals to Animats 11

T2 - From Animals to Animats 11

PB - Springer

ER -