List of Speakers

Presentation: Inference, Prediction, and Representations for Safe Navigation

Abstract: Autonomous systems and robots are becoming prevalent in our everyday lives and changing the foundations of our way of life. However, the desirable impacts of autonomy are only achievable if the underlying algorithms can handle the unique challenges humans present: People tend to defy expected behaviors and do not conform to many of the standard assumptions made in robotics. To design safe, trustworthy autonomy, we must transform how intelligent systems interact, influence, and predict human agents. In this talk, we'll discuss how inferring hidden states (e.g., traits, intent) coupled with robust prediction methods can be used to improve decision-making and control in interactive settings. Specifically, we’ll discuss recent work in goal-oriented prediction that simultaneously estimates intent and predicts the future trajectory. We'll also consider the influences and interactions between agents (humans and/or robots) in crowded settings, resulting in new structural representations for reinforcement learning that produce safe, socially-aware policies for crowd navigation. We demonstrate the effectiveness of our proposed approaches on fully equipped autonomous vehicles and mobile robots that operate among humans with ease (most of the time).

Presentation: Design Teleoperation Interfaces for Mobile Humanoid Nursing Robots

Abstract: Tele-robotic systems show promise to support the human healthcare workforce in the on-going COVID-19 crisis and during future pandemic outbreaks. Future tele-nursing robots are expected to go beyond mere tele-presence, and to perform nursing assistance tasks that involve the coordinated, remote control of manipulation, navigation, and active perception. Despite the recent advancement of robot hardware and autonomy, the desired human-robot teaming hasn't been realized because of the lack of efficient, intuitive and ergonomic teleoperation interfaces. The resulting workload and training effort also discourages the nursing workforce from adopting robots in the future workplace and during education. This talk presents our recent development of the tele-nursing robot interfaces to address these problems, including: 1) the human factor experiment that reveal the perception-action coupling and visual-haptic sensory integration in the usage of active telepresence; 2) the remote perception problems in the teleoperation of mobile manipulators to perform various navigation and loco-manipulation nursing tasks; and 3) the integration haptic and augmented reality visual feedback for assisting dexterous tele-manipulation.

Presentation: Non-Contact and Emotionally Safe Interaction of Physiotherapist Social Robot with Autistic Children for Autism Diagnosis and Treatment

Abstract: Children with autism spectrum disorder often need different physiotherapy practices to learn new skills. However, physiotherapists can be available for these children for a limited amount of time. Since studies show that autistic children positively interact with social robots, it is convincing to use these robots as physiotherapists for these children. Each child is unique regarding their abilities and needs, therefore, the robot interacting with the autistic child must be smart, diagnose the severity of autism and practice appropriately with the child. For this purpose, the robotic system needs to obtain information and sensor signals from the child. However, it is common for autistic children to suffer hyper-sensory overload. For these children, it is intolerable to wear objects around the head, hands, and other body parts. Therefore, wearable sensors are emotionally threatening for these children.

We have developed two algorithms for a social robot to interact with autistic children only based on data from a non-contact motion tracking system which makes the robot emotionally safe and useable for autistic children. The first algorithm is called segment-based online dynamic time warping (SODTW) and can diagnose autism based on the motion performance of the autistic child. The second algorithm is based on deep reinforcement learning. Using this algorithm, the robot learns from the child and adapts the speed and shape of the robot motion to generate a suitable physiotherapy motion for the child to practice. The result of these two algorithms is that the robot safely interacts with the child, adapts itself to the child, and evaluates the child’s performance over time. It is also able to save data from the child so the human physiotherapist can track the child's performance and progress.

Presentation: Fundamental Concerns in Human-Robot Co-working

Abstract: The concept of cobots and its application in physical collaboration of humans and robots have been a top buzz. Despite many commercial models becoming available, there is yet little sign of cobots being effectively used in the industries. This talk will examine several related aspects and explain some fundamental concerns that are hampering the application. The factors to be discussed include collision mechanics, psychological perception, productivity, and risk management. Such factors sit beyond the conventional theories of robotics but fundamentally challenge the application of cobots.

Presentation: Control of Unstable Physical Human-Machine Interactions: A Rider-Bikebot Example

Abstract: Human with trained motor skills can fluidly and flexibly interact with machines while smart machines can also provide motor assistance and enhancement to facilitate human’s motor skills learning. However, we currently lack theories and design tools to effectively model and tune human motor control and its interactions with machines. In this talk, I will discuss recent developments control of human motor skills through unstable physical human-machine interactions (upHMI). Rider-bikebot (i.e., bicycle-like robot) interactions is used as an upHMI paradigm to examine a sensorimotor theory for modeling of human motor control relevant to balancing motor activities. I will present a balance equilibrium manifold (BEM) concept to study how a human rider balances a bikebot while maintaining tracking a desired trajectory. A performance metric is also introduced to quantify the balance motor skills using the BEM. Extensive experiments are conducted to validate the analyses and demonstrate the balance skill metrics. Finally, I will briefly present balancing stability analysis and motor skill control of the rider-bikebot interactions.

Presentation: Toward Voice Based User Profiling in HRI

Abstract: Human-robot seamless and collaborative partnership will be a strategic component in commercial applications in the next decades. Consequently, social robotics is one of the most important and critical challenges in robot science and engineering. In this context, safety is a critical issue. Accordingly, user profiling in the physical, cognitive and social domains is crucial, and social robots should observe multimodal inputs from the human teammates. However, some mode inputs such as physiological signals require wearable sensors that may be invasive from the user point of view. Also, image processing may not be always possible depending on the operating conditions. In contrast, speech conveys a huge amount of linguistic and paralinguistic (e.g. prosody) information. Beyond voice commands to robots, speech is a window to the psychological, physical and emotional condition of humans. However, speech analysis and processing are very sensitive to noise environments (including the “cocktail party” effect), reverberation and time-varying acoustic channel that results from dynamic scenarios. In this talk we will present our recent results at the Speech Processing and Transmission Lab on robust voice based HRI in indoor reverberant environments. First, we will discuss the importance of speech based HRI in collaborative HRI. Then, the problem of time-varying acoustic channel is presented and addressed in mobile HRI by making use of deep learning and image-based beamforming. Finally, the results of our investigation on the effect in the beamforming effectiveness of robot head movement to follow the audio target are presented.