We present a controller for human-robot handovers that is automatically synthesized from high-level specifications in Signal Temporal Logic (STL). In contrast to existing controllers, this approach can provide formal guarantees on the timing of each handover phases. Using synthesis also allows end-users specify and dynamically change robot's behaviors using requirements goals constraints rather than by tuning low-level parameters. illustrate proposed replicating behavior strategies...

Deployment of Reinforcement Learning (RL) algorithms for robotics applications in the real world requires ensuring safety robot and its environment. Safe Robot RL (SRRL) is a crucial step toward achieving human-robot coexistence. In this paper, we envision human-centered SRRL framework consisting three stages: safe exploration, value alignment, collaboration. We examine research gaps these areas propose to leverage interactive behaviors SRRL. Interactive enable bi-directional information...

We present the results of two studies investigating gaze behaviors a robot receiving an object from human. Robot is important nonverbal behavior during human-robot handovers, yet prior work has only studied robots as givers. From frame-by-frame video analysis human-human we identified four receiver behaviors: gazing at giver's hand, their face, and kinds face-hand transition gazes. implemented these on arm equipped with anthropomorphic head. In Study 1, participants compared videos handover...

We evaluate the potential of Guided Policy Search (GPS), a model-based reinforcement learning (RL) method, to train robot controller for human-robot object handovers. Handovers are key competency collaborative robots and GPS could be promising approach this task, as it is data efficient does not require prior knowledge environment dynamics. However, existing uses did consider important aspects handovers, namely large spatial variations in reach locations, moving targets, generalizing over...

Shared dynamics models are important for capturing the complexity and variability inherent in Human-Robot Interaction (HRI). Therefore, learning such shared can enhance coordination adaptability to enable successful reactive interactions with a human partner. In this work, we propose novel approach latent space representation HRIs from demonstrations Mixture of Experts fashion reactively generating robot actions observations. We train Variational Autoencoder (VAE) learn motions regularized...

In a controlled experiment, participants ( n=60) competed in monotonous task with an autonomous robot for real monetary incentives. For each participant, we manipulated the robot's performance and incentive level across ten rounds. round, participant's compared to would affect their odds lottery prize. Standard economic theory predicts that people's effort will increase prize value. Furthermore, recent work behavioral economics there also be discouragement effect, stronger discouraging human...

We present a multi-sensor dataset of bimanual human-to-human object handovers. The consists 240 recordings obtained from 12 pairs participants performing handovers with 10 objects, and 120 the same unimanual 5 those objects. Each recording includes giver receiver's 13 upper-body bone position orientation trajectories, trajectories for 27 markers placed on their upper bodies, two RGB-D data streams. motion are recorded at 120Hz streams 30Hz. annotated three handover phases: reach, transfer,...

We develop and evaluate two human-robot handover controllers that allow end-users to specify timing parameters for the robot reach motion, provide feedback if cannot satisfy those constraints. End-user tuning with is a useful controller feature in settings where robots have be re-programmed varying task requirements but do not programming knowledge. The we propose are both receding-horizon differ their objective function, user specified parameters, subsequently user-interface: One uses...

Hidden Markov Models with an underlying Mixture of Gaussian structure have proven effective in learning Human-Robot Interactions from demonstrations for various interactive tasks via Regression. However, a mismatch occurs when segmenting the interaction using only observed state human compared to joint and robot. To enhance this segmentation subsequently predictive abilities such Mixture-based approaches, we take hierarchical approach by additional mixture distribution on states at...

Bimanual handovers are crucial for transferring large, deformable or delicate objects. This paper proposes a framework generating kinematically constrained human-like bimanual robot motions to ensure seamless and natural robot-to-human object handovers. We use Hidden Semi-Markov Model (HSMM) reactively generate suitable response trajectories based on the observed human partner's motion. The adapted with task space constraints accurate Results from pilot study show that our approach is...

With robots poised to enter our daily environments, they will not only need work for people, but also learn from them. An active area of investigation in the robotics, machine learning, and human-robot interaction communities is design teachable that can interactively humans. To refer these research efforts, we use umbrella term Human-Interactive Robot Learning (HIRL). In last 2 years began consolidating what defines HIRL terms long, medium, short-term problems different contribute those...

This paper explores active sensing strategies that employ vision-based tactile sensors for robotic perception and classification of fabric textures. We formalize the sampling problem in context recognition provide an implementation information-theoretic exploration based on minimizing predictive entropy variance probabilistic models. Through ablation studies human experiments, we investigate which components are crucial quick reliable texture recognition. Along with strategies, evaluate...

Training robot policies in simulation is becoming increasingly popular; nevertheless, a precise, reliable, and easy-to-use tactile simulator for contact-rich manipulation tasks still missing. To close this gap, we develop TacEx -- modular framework. We embed state-of-the-art soft-body contacts named GIPC vision-based simulators Taxim FOTS into Isaac Sim to achieve robust plausible of the visuotactile sensor GelSight Mini. implement several Lab environments Reinforcement Learning (RL)...

Planetary rovers offer promising alternatives for conducting in-situ experiments on other planets. Considering the huge costs of such exploration missions, it becomes necessary to design an optimal mobility system configuration meet mission goals. In this paper, a constrained optimization procedure is presented obtain parameters rocker-bogie system. More recently, researchers have proposed idea using reconfigurable wheels, which would be able change their shape according terrain, improve...

This work aims at asymptotically accurate dimensional reduction of non-linear multi-functional film-fabric laminates having specific application in design envelopes for High Altitude Airships (HAA). The laminate airship envelope consists a woven fabric core coated with thin films on each face. These provide UV protection and Helium leakage prevention, while the provides required structural strength. problem is both geometrically materially non-linear. To incorporate geometric non-linearity,...

Deployment of Reinforcement Learning (RL) algorithms for robotics applications in the real world requires ensuring safety robot and its environment. Safe Robot RL (SRRL) is a crucial step towards achieving human-robot coexistence. In this paper, we envision human-centered SRRL framework consisting three stages: safe exploration, value alignment, collaboration. We examine research gaps these areas propose to leverage interactive behaviors SRRL. Interactive enable bi-directional information...