George Velentzas

George Velentzas

Auto-generated profile, based on publicly available information

George Velentzas

School of Electrical and Computer Engineering, National Technical University of Athens, Greece

Research Focus & Metrics

Academic Career Length

10Years (2017-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
5 July 2018 in IEEE Transactions on Cognitive and Developmental Systems
Verified
Journal Article
Open access
MKMehdi Khamassi
Mehdi Khamassi
GVGeorge Velentzas
George Velentzas
TTTheodore Tsitsimis
Theodore Tsitsimis
Costas S. TZAFESTASCostas S. TZAFESTAS
Costas S. TZAFESTAS
Dynamic uncontrolled human-robot interactions (HRI) require robots to be able to adapt to changes in the human’s behavior and intentions. Among relevant signals, non-verbal cues such as the human’s gaze can provide the robot with important information about the human’s current engagement in the task, and whether the robot should continue its current behavior or not. However, robot reinforcement learning (RL) abilities to adapt to these non-verbal cues are still underdeveloped. Here we propose an active exploration algorithm for RL during HRI where the reward function is the weighted sum of the human’s current engagement and variations of this engagement. We use a parameterized action space where a meta-learning algorithm is applied to simultaneously tune the exploration in discrete action space (e.g. moving an object) and in the space of continuous characteristics of movement (e.g. velocity, direction, strength, expressivity). We first show that this algorithm reaches state-of-the-art performance in the non-stationary multi-armed bandit paradigm. We then apply it to a simulated HRI task, and show that it outperforms continuous parameterized RL with either passive or active exploration based on different existing methods. We finally test the performance in a more realistic test of the same HRI task, where a practical approach is followed to estimate human engagement through visual cues of the head pose. The algorithm can detect and adapt to perturbations in human engagement with different durations. Altogether, these results suggest a novel efficient and robust framework for robot learning during dynamic HRI scenarios.

1 April 2017
Verified
Conference Paper
Open access
MKMehdi Khamassi
Mehdi Khamassi
GVGeorge Velentzas
George Velentzas
TTTheodore Tsitsimis
Theodore Tsitsimis
Costas S. TZAFESTASCostas S. TZAFESTAS
Costas S. TZAFESTAS
Online model-free reinforcement learning (RL) methods with continuous actions are playing a prominent role when dealing with real-world applications such as Robotics. However, when confronted to non-stationary environments, these methods crucially rely on an exploration-exploitation trade-off which is rarely dynamically and automatically adjusted to changes in the environment. Here we propose an active exploration algorithm for RL in structured (parameterized) continuous action space. This framework deals with a set of discrete actions, each of which is parameterized with continuous variables. Discrete exploration is controlled through a Boltzmann softmax function with an inverse temperature β parameter. In parallel, a Gaussian exploration is applied to the continuous action parameters. We apply a meta-learning algorithm based on the comparison between variations of short-term and long-term reward running averages to simultaneously tune β and the width of the Gaussian distribution from which continuous action parameters are drawn. We first show that this algorithm reaches state-of-the-art performance in the non-stationary multi-armed bandit paradigm, while also being generalizable to continuous actions and multi-step tasks. We then apply it to a simulated human-robot interaction task, and show that it outperforms continuous parameterized RL both without active exploration and with active exploration based on uncertainty variations measured by a Kalman-Q-learning algorithm.

1 July 2018 in Paladyn
Verified
Journal Article
Open access
PDF file
GVGeorge Velentzas
George Velentzas
TTTheodore Tsitsimis
Theodore Tsitsimis
IRIñaki Rañó
Iñaki Rañó
Costas S. TZAFESTASCostas S. TZAFESTAS
Costas S. TZAFESTAS
MKMehdi Khamassi
Mehdi Khamassi
Using assistive robots for educational applications requires robots to be able to adapt their behavior specifically for each child with whom they interact.Among relevant signals, non-verbal cues such as the child’s gaze can provide the robot with important information about the child’s current engagement in the task, and whether the robot should continue its current behavior or not. Here we propose a reinforcement learning algorithm extended with active state-specific exploration and show its applicability to child engagement maximization as well as more classical tasks such as maze navigation. We first demonstrate its adaptive nature on a continuous maze problem as an enhancement of the classic grid world. There, parameterized actions enable the agent to learn single moves until the end of a corridor, similarly to “options” but without explicit hierarchical representations.We then apply the algorithm to a series of simulated scenarios, such as an extended Tower of Hanoi where the robot should find the appropriate speed of movement for the interacting child, and to a pointing task where the robot should find the child-specific appropriate level of expressivity of action. We show that the algorithm enables to cope with both global and local non-stationarities in the state space while preserving a stable behavior in other stationary portions of the state space. Altogether, these results suggest a promising way to enable robot learning based on non-verbal cues and the high degree of non-stationarities that can occur during interaction with children.