Browsing by Author "Asada, M."

Now showing 1 - 11 of 11

Metadata only
Bimanual rope manipulation skill synthesis through context dependent correction policy learning from human demonstration
(IEEE, 2023) Akbulut, B.; Girgin, T.; Mehrabi, Arash; Asada, M.; Ugur, E.; Öztop, Erhan; Computer Science; ÖZTOP, Erhan; Mehrabi, Arash
Learning from demonstration (LfD) with behavior cloning is attractive for its simplicity; however, compounding errors in long and complex skills can be a hindrance. Considering a target skill as a sequence of motor primitives is helpful in this respect. Then the requirement that a motor primitive ends in a state that allows the successful execution of the subsequent primitive must be met. In this study, we focus on this problem by proposing to learn an explicit correction policy when the expected transition state between primitives is not achieved. The correction policy is learned via behavior cloning by the use of Conditional Neural Motor Primitives (CNMPs) that can generate correction trajectories in a context-dependent way. The advantage of the proposed system over learning the complete task as a single action is shown with a table-top setup in simulation, where an object has to be pushed through a corridor in two steps. Then, the applicability of the proposed method to bi-manual knotting in the real world is shown by equipping an upper-body humanoid robot with the skill of making knots over a bar in 3D space.
Metadata only
Combined weight and density bounds on the polynomial threshold function representation of Boolean functions
(Elsevier, 2022-08) Öztop, Erhan; Asada, M.; Computer Science; ÖZTOP, Erhan
In an earlier report it was shown that an arbitrary n-variable Boolean function f can be represented as a polynomial threshold function (PTF) with 0.75×2n or less number of monomials. In this report, we derive an upper bound on the absolute value of the (integer) weights of a PTF that represents f and still obeys the aforementioned density bound. To our knowledge this provides the best combined bound on the PTF density (number of monomials) and PTF weight (sum of the coefficient magnitudes) of general Boolean functions. For the special case of bent functions, it is found that any n-variable bent function can be represented with integer coefficients less than or equal to 2n with density no more than 0.75×2n, and for the case of m-sparse Boolean functions that are almost constant except for small (m≪2n) number of variable assignments, it is shown that they can be represented with small weight PTFs with density at most m+2n−1. In addition, tight PTF weight bounds with conformance to the density bound of 0.75×2n are numerically obtained for the general Boolean functions up to 6 variables.
Metadata only
Context based echo state networks for robot movement primitives
(IEEE, 2023) Amirshirzad, Negin; Asada, M.; Öztop, Erhan; Computer Science; ÖZTOP, Erhan; Amirshirzad, Negin
Reservoir Computing, in particular Echo State Networks (ESNs) offer a lightweight solution for time series representation and prediction. An ESN is based on a discrete time random dynamical system that is used to output a desired time series with the application of a learned linear readout weight vector. The simplicity of the learning suggests that an ESN can be used as a lightweight alternative for movement primitive representation in robotics. In this study, we explore this possibility and develop Context-based Echo State Networks (CESNs), and demonstrate their applicability to robot movement generation. The CESNs are designed for generating joint or Cartesian trajectories based on a user definable context input. The context modulates the dynamics represented by the ESN involved. The linear read-out weights then can pick up the context-dependent dynamics for generating different movement patterns for different contexts. To achieve robust movement execution and generalization over unseen contexts, we introduce a novel data augmentation mechanism for ESN training. We show the effectiveness of our approach in a learning from demonstration setting. To be concrete, we teach the robot reaching and obstacle avoidance tasks in simulation and in real-world, which shows that the developed system, CESN provides a lightweight movement primitive representation system that facilitate robust task execution with generalization ability for unseen seen contexts, including extrapolated ones.
Metadata only
An ecologically valid reference frame for perspective invariant action recognition
(IEEE, 2021) Bayram, Berkay; Uğur, E.; Asada, M.; Öztop, Erhan; Computer Science; ÖZTOP, Erhan; Bayram, Berkay
In robotics, objects and body parts can be represented in various coordinate frames to ease computation. In biological systems, body or body part centered coordinate frames have been proposed as possible reference frames that the brain uses for interacting with the environment. Coordinate transformations are standard tools in robotics and can facilitate perspective invariant action recognition and action prediction based on observed actions of other agents. Although it is known that human adults can do explicit coordinate transformations, it is not clear whether this capability is used for recognizing and understanding the actions of others. Mirror neurons, found in the ventral premotor cortex of macaque monkeys, seem to undertake action understanding in a perspective invariant way, which may rely on lower level perceptual mechanisms. To this end, in this paper we propose a novel reference frame that is ecologically plausible and can sustain basic action understanding and mirror function. We demonstrate the potential of this representation by simulation of an upper body humanoid robot with an action repertoire consisting of push, poke, move-away and bring-to-mouth actions. The simulation experiments indicate that the representation is suitable for action recognition and effect prediction in a perspective invariant way, and thus can be deployed as an artificial mirror system for robotic applications.
Metadata only
Forming robot trust in heterogeneous agents during a multimodal interactive game
(IEEE, 2022) Kırtay, M.; Öztop, Erhan; Kuhlen, A. K.; Asada, M.; Hafner, V. V.; Computer Science; ÖZTOP, Erhan
This study presents a robot trust model based on cognitive load that uses multimodal cues in a learning setting to assess the trustworthiness of heterogeneous interaction partners. As a test-bed, we designed an interactive task where a small humanoid robot, Nao, is asked to perform a sequential audio-visual pattern recall task while minimizing its cognitive load by receiving help from its interaction partner, either a robot, Pepper, or a human. The partner displayed one of three guiding strategies, reliable, unreliable, or random. The robot is equipped with two cognitive modules: a multimodal auto-associative memory and an internal reward module. The former represents the multimodal cognitive processing of the robot and allows a 'cognitive load' or 'cost' to be assigned to the processing that takes place, while the latter converts the cognitive processing cost to an internal reward signal that drives the cost-based behavior learning. Here, the robot asks for help from its interaction partner when its action leads to a high cognitive load. Then the robot receives an action suggestion from the partner and follows it. After performing interactive experiments with each partner, the robot uses the cognitive load yielded during the interaction to assess the trustworthiness of the partners -i.e., it associates high trustworthiness with low cognitive load. We then give a free choice to the robot to select the trustworthy interaction partner to perform the next task. Our results show that, overall, the robot selects partners with reliable guiding strategies. Moreover, the robot's ability to identify a trustworthy partner was unaffected by whether the partner was a human or a robot.
Metadata only
High-level features for resource economy and fast learning in skill transfer
(Taylor & Francis, 2022) Ahmetoglu, A.; Uğur, E.; Asada, M.; Öztop, Erhan; Computer Science; ÖZTOP, Erhan
Abstraction is an important aspect of intelligence which enables agents to construct robust representations for effective and efficient decision making. Although, deep neural networks are proven to be effective learning systems due to their ability to form increasingly complex abstractions at successive layers these abstractions are mostly distributed over many neurons, making the re-use of a learned skill costly and blind to the insights that can be obtained on the emergent representations. For avoiding designer bias and unsparing resource use, we propose to exploit neural response dynamics to form compact representations to use in skill transfer. For this, we consider two competing methods based on (1) maximum information compression principle and (2) the notion that abstract events tend to generate slowly changing signals, and apply them to the neural signals generated during task execution. To be concrete, in our simulation experiments, we either apply principal component analysis (PCA) or slow feature analysis (SFA) on the signals collected from the last hidden layer of a deep neural network while it performs a source task, and use these features for skill transfer in a new, target, task. We then compare the generalization and learning performance of these alternatives with the baselines of skill transfer with full layer output and no-transfer settings. Our experimental results on a simulated tabletop robot arm navigation task show that units that are created with SFA are the most successful for skill transfer. SFA as well as PCA, incur less resources compared to usual skill transfer where full layer outputs are used in the new task learning, whereby many units formed show a localized response reflecting end-effector-obstacle-goal relations. Finally, SFA units with the lowest eigenvalues resemble symbolic representations that highly correlate with high-level features such as joint angles and end-effector position which might be thought of as precursors for fully symbolic systems.
Open Access
Imitation and mirror systems in robots through Deep Modality Blending Networks
(Elsevier, 2022-02) Seker, M. Y.; Ahmetoglu, A.; Nagai, Y.; Asada, M.; Öztop, Erhan; Ugur, E.; Computer Science; ÖZTOP, Erhan
Learning to interact with the environment not only empowers the agent with manipulation capability but also generates information to facilitate building of action understanding and imitation capabilities. This seems to be a strategy adopted by biological systems, in particular primates, as evidenced by the existence of mirror neurons that seem to be involved in multi-modal action understanding. How to benefit from the interaction experience of the robots to enable understanding actions and goals of other agents is still a challenging question. In this study, we propose a novel method, deep modality blending networks (DMBN), that creates a common latent space from multi-modal experience of a robot by blending multi-modal signals with a stochastic weighting mechanism. We show for the first time that deep learning, when combined with a novel modality blending scheme, can facilitate action recognition and produce structures to sustain anatomical and effect-based imitation capabilities. Our proposed system, which is based on conditional neural processes, can be conditioned on any desired sensory/motor value at any time step, and can generate a complete multi-modal trajectory consistent with the desired conditioning in one-shot by querying the network for all the sampled time points in parallel avoiding the accumulation of prediction errors. Based on simulation experiments with an arm-gripper robot and an RGB camera, we showed that DMBN could make accurate predictions about any missing modality (camera or joint angles) given the available ones outperforming recent multimodal variational autoencoder models in terms of long-horizon high-dimensional trajectory predictions. We further showed that given desired images from different perspectives, i.e. images generated by the observation of other robots placed on different sides of the table, our system could generate image and joint angle sequences that correspond to either anatomical or effect-based imitation behavior. To achieve this mirror-like behavior, our system does not perform a pixel-based template matching but rather benefits from and relies on the common latent space constructed by using both joint and image modalities, as shown by additional experiments. Moreover, we showed that mirror learning (in our system) does not only depend on visual experience and cannot be achieved without proprioceptive experience. Our experiments showed that out of ten training scenarios with different initial configurations, the proposed DMBN model could achieve mirror learning in all of the cases where the model that only uses visual information failed in half of them. Overall, the proposed DMBN architecture not only serves as a computational model for sustaining mirror neuron-like capabilities, but also stands as a powerful machine learning architecture for high-dimensional multi-modal temporal data with robust retrieval capabilities operating with partial information in one or multiple modalities.
Metadata only
Interplay between neural computational energy and multimodal processing in robot-robot interaction
(IEEE, 2023) Kırtay, M.; Hafner, V. V.; Asada, M.; Öztop, Erhan; Computer Science; ÖZTOP, Erhan
Multimodal learning is an active research area that is gaining importance in human-robot interaction. Despite the obvious benefit of levering multiple sensors for perceiving the world, its neural computational cost has not been addressed in robotics, especially in Robot-Robot Interaction (RRI). This study addresses the role of computational cost in multimodal processing by considering robot-robot interaction in a sequential multimodal memory recall task. In this setting, the learner (Nao) robot receives auditory-only, visual-only, or audio-visual information from an instructor (Pepper) robot and the environment regarding previously learned memory items. The goal of the learner robot is to perform the interactive task with as low as possible neural computational cost. The learner robot has two cognitive modules: a multimodal auto-associative network that stands for the perceptual-cognitive processing of the robot and an internal reward mechanism that monitors the changes in neural energy incurred for two consecutive steps by the processing of the attended stimuli. The reward computed is used to build an action policy for minimizing the neural energy consumption over the sequential memory recall task. The experimental results show that having access to both auditory and visual information is beneficial not only for better memory recall but also for minimizing the cost of neural computation.
Metadata only
Modeling robot trust based on emergent emotion in an interactive task
(IEEE, 2021) Kırtay, M.; Öztop, Erhan; Asada, M.; Hafner, V. V.; Computer Science; ÖZTOP, Erhan
Trust is an essential component in human-human and human-robot interactions. The factors that play potent roles in these interactions have been an attractive issue in robotics. However, the studies that aim at developing a computational model of robot trust in interaction partners remain relatively limited. In this study, we extend our emergent emotion model to propose that the robot's trust in the interaction partner (i.e., trustee) can be established by the effect of the interactions on the computational energy budget of the robot (i.e., trustor). To be concrete, we show how high-level emotions (e.g., wellbeing) of an agent can be modeled by the computational cost of perceptual processing (e.g., visual stimulus processing for visual recalling) in a decision-making framework. To realize this approach, we endow the Pepper humanoid robot with two modules: an auto-associative memory that extracts the required computational energy to perform a visual recalling, and an internal reward mechanism guiding model-free reinforcement learning to yield computational energy cost-aware behaviors. With this setup, the robot interacts with online instructors with different guiding strategies, namely reliable, less reliable, and random. Through interaction with the instructors, the robot associates the cumulative reward values based on the cost of perceptual processing to evaluate the instructors and determine which one should be trusted. Overall the results indicate that the robot can differentiate the guiding strategies of the instructors. Additionally, in the case of free choice, the robot trusts the reliable one that increases the total reward - and therefore reduces the required computational energy (cognitive load)- to perform the next task.
Metadata only
Multimodal reinforcement learning for partner specific adaptation in robot-multi-robot interaction
(IEEE, 2022) Kırtay, M.; Hafner, V. V.; Asada, M.; Kuhlen, A. K.; Öztop, Erhan; Computer Science; ÖZTOP, Erhan
Successful and efficient teamwork requires knowledge of the individual team members' expertise. Such knowledge is typically acquired in social interaction and forms the basis for socially intelligent, partner-Adapted behavior. This study aims to implement this ability in teams of multiple humanoid robots. To this end, a humanoid robot, Nao, interacted with three Pepper robots to perform a sequential audio-visual pattern recall task that required integrating multimodal information. Nao outsourced its decisions (i.e., action selections) to its robot partners to perform the task efficiently in terms of neural computational cost by applying reinforcement learning. During the interaction, Nao learned its partners' specific expertise, which allowed Nao to turn for guidance to the partner who has the expertise corresponding to the current task state. The cognitive processing of Nao included a multimodal auto-Associative memory that allowed the determination of the cost of perceptual processing (i.e., cognitive load) when processing audio-visual stimuli. In turn, the processing cost is converted into a reward signal by an internal reward generation module. In this setting, the learner robot Nao aims to minimize cognitive load by turning to the partner whose expertise corresponds to a given task state. Overall, the results indicate that the learner robot discovers the expertise of partners and exploits this information to execute its task with low neural computational cost or cognitive load.
Metadata only
Trust me! I am a robot: an affective computational account of scaffolding in robot-robot interaction
(IEEE, 2021-08-08) Kırtay, M.; Öztop, Erhan; Asada, M.; Hafner, V. V:; Computer Science; ÖZTOP, Erhan
Forming trust in a biological or artificial interaction partner that provides reliable strategies and employing the learned strategies to scaffold another agent are critical problems that are often addressed separately in human-robot and robot-robot interaction studies. In this paper, we provide a unified approach to address these issues in robot-robot interaction settings. To be concrete, we present a trust-based affective computational account of scaffolding while performing a sequential visual recalling task. In that, we endow the Pepper humanoid robot with cognitive modules of auto-associative memory and internal reward generation to implement the trust model. The former module is an instance of a cognitive function with an associated neural cost determining the cognitive load of performing visual memory recall. The latter module uses this cost to generate an internal reward signal to facilitate neural cost-based reinforcement learning (RL) in an interactive scenario involving online instructors with different guiding strategies: reliable, less-reliable, and random. These cognitive modules allow the Pepper robot to assess the instructors based on the average cumulative reward it can collect and choose the instructor that helps reduce its cognitive load most as the trustworthy one. After determining the trustworthy instructor, the Pepper robot is recruited to be a caregiver robot to guide a perceptually limited infant robot (i.e., the Nao robot) that performs the same task. In this setting, we equip the Pepper robot with a simple theory of mind module that learns the state-action-reward associations by observing the infant robot's behavior and guides the learning of the infant robot, similar to when it went through the online agent-robot interactions. The experiment results on this robot-robot interaction scenario indicate that the Pepper robot as a caregiver leverages the decision-making policies - obtained by interacting with the trustworthy instructor- to guide the infant robot to perform the same task efficiently. Overall, this study suggests how robotic-trust can be grounded in human-robot or robot-robot interactions based on cognitive load, and be used as a mechanism to choose the right scaffolding agent for effective knowledge transfer.