Using echo state networks for robot navigation behavior acquisition

Abstract

Abstract-Robot Behavior Learning by Demonstration deals with the ability for a robot to learn a behavior from one or several demonstrations provided by a human teacher, possibly through tele-operation or imitation. This implies controllers that can address both (1) the feature selection problem related to a great amount of mostly irrelevant sensory data and (2) dealing with temporal sequences of demonstrations. Echo State Networks [10] have been proposed recently for time series prediction and have been shown to perform remarkably well on this kind of data. In this paper, we introduce ESN to robot behavior acquisition in the scope of a mobile robot performing navigation tasks. ESN actually show comparable and even better performance with that of other algorithms from the literature in similar experimental conditions. Moreover, some properties regarding dynamics of ESN in the context of learning by demonstration are investigated.

Trajectory learning and reproduction for differential drive mobile robots based on GMM/HMM and dynamic time warping using learning from demonstration framework

“…Another research result based on BBR interpretation of LfD is provided in Narayanan et al (2011), where authors teach the mobile robot to navigate in an unknown environment using neural networks to learn specific behaviors, based on human demonstration and sensor information provided by omnidirectional camera. Similar approach can be seen in Hartland and Bredeche (2007) where authors develop the echo state recurrent neural network and teach the mobile robot to navigate using the LfD approach.…”

“…For example, the field is rich in regression techniques like locally weighted regression (Atkeson and Schaal, 1997;Argall et al, 2011), swarm intelligence (Muñoz et al 2014), neural networks (Pomerleau, 1995;Hartland and Bredeche, 2007;Suleman and Awais, 2011), ARMAX and NARMAX system identification (Nehmzow et al, 2007(Nehmzow et al, , 2010, probabilistic methods (Abbeel et al, 2010;Calinon et al, 2010;Konidaris et al, 2011), or mixtures of applied methods (Narayanan et al, 2011;Vakanski et al, 2012). This feature of the LfD represents a significant advantage that enabled a variety of robotic application.…”

On Learning Navigation Behaviors for Small Mobile Robots With Reservoir Computing Architectures

“…In [31], an ESN is used to model behavior acquisition by demonstration for a Khepera mobile robot using an 8x6 color image as input to the network. They train the ESN to perform a sequence of reactive behaviors (find and reach target), which actually do not require the dynamic properties of the reservoir since their results show that the same performance can be achieved if the recurrent connections from the reservoir are removed.…”

Computing with Spiking Neuron Networks

“…SNN-based systems also develop increasingly in the area of robotics, where fast processing is a key issue [104,163,44,43,126,50], from wheels to wings, or legged locomotion. The special abilities of SNNs for fast computing transient temporal patterns make them on first line for designing efficient systems in the area of autonomous robotics.…”