XBG: End-to-end Imitation Learning for Autonomous Behaviour in Human-Robot Interaction and Collaboration

This paper presents XBG (eXteroceptive Behaviour Generation), a multimodal end-to-end Imitation Learning (IL) system for a whole-body autonomous humanoid robot used in real-world Human-Robot Interaction (HRI) scenarios. The main contribution of this paper is an architecture for learning HRI behaviours using a data-driven approach. Through teleoperation, a diverse dataset is collected, comprising demonstrations across multiple HRI scenarios, including handshaking, handwaving, payload reception, walking, and walking with a payload. After synchronizing, filtering, and transforming the data, different Deep Neural Networks (DNN) models are trained. The final system integrates different modalities comprising exteroceptive and proprioceptive sources of information to provide the robot with an understanding of its environment and its own actions. The robot takes sequence of images (RGB and depth) and joints state information during the interactions and then reacts accordingly, demonstrating learned behaviours. By fusing multimodal signals in time, we encode new autonomous capabilities into the robotic platform, allowing the understanding of context changes over time. The models are deployed on ergoCub, a real-world humanoid robot, and their performance is measured by calculating the success rate of the robot's behaviour under the mentioned scenarios.

Codesign of Humanoid Robots for Ergonomy Collaboration with Multiple Humans via Genetic Algorithms and Nonlinear Optimization

Ergonomics is a key factor to consider when designing control architectures for effective physical collaborations between humans and humanoid robots. In contrast, ergonomic indexes are often overlooked in the robot design phase, which leads to suboptimal performance in physical human-robot interaction tasks. This paper proposes a novel methodology for optimizing the design of humanoid robots with respect to ergonomic indicators associated with the interaction of multiple agents. Our approach leverages a dynamic and kinematic parameterization of the robot link and motor specifications to seek for optimal robot designs using a bilevel optimization approach. Specifically, a genetic algorithm first generates robot designs by selecting the link and motor characteristics. Then, we use nonlinear optimization to evaluate interaction ergonomy indexes during collaborative payload lifting with different humans and weights. To assess the effectiveness of our approach, we compare the optimal design obtained using bilevel optimization against the design obtained using nonlinear optimization. Our results show that the proposed approach significantly improves ergonomics in terms of energy expenditure calculated in two reference scenarios involving static and dynamic robot motions. We plan to apply our methodology to drive the design of the ergoCub2 robot, a humanoid intended for optimal physical collaboration with humans in diverse environments

A Flexible MATLAB/Simulink Simulator for Robotic Floating-base Systems in Contact with the Ground

Physics simulators are widely used in robotics fields, from mechanical design to dynamic simulation, and controller design. This paper presents an open-source MATLAB/Simulink simulator for rigid-body articulated systems, including manipulators and floating-base robots. Thanks to MATLAB/Simulink features like MATLAB system classes and Simulink function blocks, the presented simulator combines a programmatic and block-based approach, resulting in a flexible design in the sense that different parts, including its physics engine, robot-ground interaction model, and state evolution algorithm are simply accessible and editable. Moreover, through the use of Simulink dynamic mask blocks, the proposed simulation framework supports robot models integrating open-chain and closed-chain kinematics with any desired number of links interacting with the ground. The simulator can also integrate second-order actuator dynamics. Furthermore, the simulator benefits from a one-line installation and an easy-to-use Simulink interface.

Whole-Body Human Kinematics Estimation using Dynamical Inverse Kinematics and Contact-Aided Lie Group Kalman Filter

Full-body motion estimation of a human through wearable sensing technologies is challenging in the absence of position sensors. This paper contributes to the development of a model-based whole-body kinematics estimation algorithm using wearable distributed inertial and force-torque sensing. This is done by extending the existing dynamical optimization-based Inverse Kinematics (IK) approach for joint state estimation, in cascade, to include a center of pressure-based contact detector and a contact-aided Kalman filter on Lie groups for floating base pose estimation. The proposed method is tested in an experimental scenario where a human equipped with a sensorized suit and shoes performs walking motions. The proposed method is demonstrated to obtain a reliable reconstruction of the whole-body human motion.

An Experimental Comparison of Floating Base Estimators for Humanoid Robots with Flat Feet

Extended Kalman filtering is a common approach to achieve floating base estimation of a humanoid robot. These filters rely on measurements from an Inertial Measurement Unit (IMU) and relative forward kinematics for estimating the base position-and-orientation and its linear velocity along with the augmented states of feet position-and-orientation, thus giving them their name, flat-foot filters. However, the availability of only partial measurements often poses the question of consistency in the filter design. In this paper, we perform an experimental comparison of state-of-the-art flat-foot filters based on the representation choice of state, observation, matrix Lie group error and system dynamics evaluated for filter consistency and trajectory errors. The comparison is performed over simulated and real-world experiments conducted on the iCub humanoid platform.

Efficient Geometric Linearization of Moving-Base Rigid Robot Dynamics

The linearization of the equations of motion of a robotics system about a given state-input trajectory, including a controlled equilibrium state, is a valuable tool for model-based planning, closed-loop control, gain tuning, and state estimation. Contrary to the case of fixed based manipulators with prismatic or rotary joints, the state space of moving-base robotic systems such as humanoids, quadruped robots, or aerial manipulators cannot be globally parametrized by a finite number of independent coordinates. This impossibility is a direct consequence of the fact that the state of these systems includes the system's global orientation, formally described as an element of the special orthogonal group SO(3). As a consequence, obtaining the linearization of the equations of motion for these systems is typically resolved, from a practical perspective, by locally parameterizing the system's attitude by means of, e.g., Euler or Cardan angles. This has the drawback, however, of introducing artificial parameterization singularities and extra derivative computations. In this contribution, we show that it is actually possible to define a notion of linearization that does not require the use of a local parameterization for the system's orientation, obtaining a mathematically elegant, recursive, and singularity-free linearization for moving-based robot systems. Recursiveness, in particular, is obtained by proposing a nontrivial modification of existing recursive algorithms to allow for computations of the geometric derivatives of the inverse dynamics and the inverse of the mass matrix of the robotic system. The correctness of the proposed algorithm is validated by means of a numerical comparison with the result obtained via geometric finite difference.

iCub3 Avatar System

We present an avatar system that enables a human operator to visit a remote location via iCub3, a new humanoid robot developed at the Italian Institute of Technology (IIT) paving the way for the next generation of the iCub platforms. On the one hand, we present the humanoid iCub3 that plays the role of the robotic avatar. Particular attention is paid to the differences between iCub3 and the classical iCub humanoid robot. On the other hand, we present the set of technologies of the avatar system at the operator side. They are mainly composed of iFeel, namely, IIT lightweight non-invasive wearable devices for motion tracking and haptic feedback, and of non-IIT technologies designed for virtual reality ecosystems. Finally, we show the effectiveness of the avatar system by describing a demonstration involving a realtime teleoperation of the iCub3. The robot is located in Venice, Biennale di Venezia, while the human operator is at more than 290km distance and located in Genoa, IIT. Using a standard fiber optic internet connection, the avatar system transports the operator locomotion, manipulation, voice, and face expressions to the iCub3 with visual, auditory, haptic and touch feedback.

Online Non-linear Centroidal MPC for Humanoid Robot Locomotion with Step Adjustment

This paper presents a Non-Linear Model Predictive Controller for humanoid robot locomotion with online step adjustment capabilities. The proposed controller considers the Centroidal Dynamics of the system to compute the desired contact forces and torques and contact locations. Differently from bipedal walking architectures based on simplified models, the presented approach considers the reduced centroidal model, thus allowing the robot to perform highly dynamic movements while keeping the control problem still treatable online. We show that the proposed controller can automatically adjust the contact location both in single and double support phases. The overall approach is then tested with a simulation of one-leg and two-leg systems performing jumping and running tasks, respectively. We finally validate the proposed controller on the position-controlled Humanoid Robot iCub. Results show that the proposed strategy prevents the robot from falling while walking and pushed with external forces up to 40 Newton for 1 second applied at the robot arm.

DILIGENT-KIO: A Proprioceptive Base Estimator for Humanoid Robots using Extended Kalman Filtering on Matrix Lie Groups

This paper presents a contact-aided inertial-kinematic floating base estimation for humanoid robots considering an evolution of the state and observations over matrix Lie groups. This is achieved through the application of a geometrically meaningful estimator which is characterized by concentrated Gaussian distributions. The configuration of a floating base system like a humanoid robot usually requires the knowledge of an additional six degrees of freedom which describes its base position-and-orientation. This quantity usually cannot be measured and needs to be estimated. A matrix Lie group, encapsulating the position-and-orientation and linear velocity of the base link, feet positions-and-orientations and Inertial Measurement Units' biases, is used to represent the state while relative positions-and-orientations of contact feet from forward kinematics are used as observations. The proposed estimator exhibits fast convergence for large initialization errors owing to choice of uncertainty parametrization. An experimental validation is done on the iCub humanoid platform.

On the Emergence of Whole-body Strategies from Humanoid Robot Push-recovery Learning

Balancing and push-recovery are essential capabilities enabling humanoid robots to solve complex locomotion tasks. In this context, classical control systems tend to be based on simplified physical models and hard-coded strategies. Although successful in specific scenarios, this approach requires demanding tuning of parameters and switching logic between specifically-designed controllers for handling more general perturbations. We apply model-free Deep Reinforcement Learning for training a general and robust humanoid push-recovery policy in a simulation environment. Our method targets high-dimensional whole-body humanoid control and is validated on the iCub humanoid. Reward components incorporating expert knowledge on humanoid control enable fast learning of several robust behaviors by the same policy, spanning the entire body. We validate our method with extensive quantitative analyses in simulation, including out-of-sample tasks which demonstrate policy robustness and generalization, both key requirements towards real-world robot deployment.