Program

Abstract: Object handovers, a seemingly straightforward action, involve complex interplays of predictive and reactive control mechanisms in both partners. Non-verbal communication is a fundamental requirement for these mechanisms to work. In this work, we show that in an agent-human object handover, the human receiver can use passer kinematic cues to predict the transported object's properties, such as weight, and preemptively adapt the grasping strategy towards them. In particular, we show that when the agent's pick-lift-transport movement is correlated to the object weight, humans can interpret this cue and produce proportional anticipatory grip forces before release of the object. This adaptation can be learned even when objects are presented in a random order and is further improved with the repeated presentation of the same agent trajectory-object weight pair. Conversely,  the anticipatory load force scaling reduces when the object weight is learned. We attribute this reduction to a delayed load force generation of participants, which resulted in improved synchronization with the agent's object release. The outcome of this study contributes to a better understanding of non-verbal cues in handover tasks and enables more transparent and efficient physical robot-human interactions.

Abstract: Inspired by the natural world, where certain animals utilize tails for balance, this paper introduces a Supernumerary Robotic Limb (SRLs) in the form of a wearable robotic tail designed to support human balance and posture during pick-and-place tasks. In industries such as logistics and construction, humans frequently perform tasks including loading and unloading shelves, palatalisation, and manufacturing on construction lines. These repetitive tasks can include lifting and moving heavy objects which, over time, can lead to strains and injuries. This can result in reduced productivity and potentially lost jobs as workers age. To address this issue, we have developed a wearable robotic tail designed to support balance and posture during load-carrying tasks. We tested this device on a healthy participant performing pick-and-place tasks. The participant's Centre of Pressure in the sagittal plane was monitored, along with the muscle activity of the Erector Spinae and Gastrocnemius muscles. Our observations revealed a 57\% reduction in forward displacement of the Centre of Pressure and significant reductions in muscle activity. These results demonstrate the potential of the device to reduce body strain, thereby helping workers maintain their health and productivity.

Abstract: Falls among older adults pose significant health risks, leading to severe injuries and substantial economic costs, with approximately 684,000 deaths and 37.3 million medical treatments annually. This study introduces a novel fall detection algorithm using machine learning, validated with data from the Milbotix smart sock capturing acceleration data and XSens, a motion capture system. Thirty healthy participants wearing XSens and smart socks participated in the data collection. The collected data will be labelled and pre-processed for machine learning model training. By integrating these data sources, our algorithm achieves accuracies of 98.09% for the Smart Sock and 99.19% for the XSens, demonstrating improved accuracy in fall detection. Additionally, based on a single position sensor, the sock's portability and ease of setup significantly enhance user experience over existing methods.

Abstract: This study introduces a camera-enabled smart walker system with real-time auditory feedback to monitor and correct user posture. Using Google Mediapipe's Pose Detection and a custom robotic platform, the system accurately classifies eight distinct postures. In two case studies with healthy participants during walker-assisted gait, the system achieved over 80\% accuracy for each posture category, including 98.15\% for standing and 98.45\% for lifting hand movements. A third case study, using a dataset from 22 participants, yielded an overall accuracy of 59.37\%, highlighting challenges in generalisations across multiple users. This proof-of-concept demonstrates the potential for integrating advanced technology with traditional mobility aids to improve safety and quality of life for the elderly. Future work will focus on epitomising posture detection algorithms and exploring alternative vision models to enhance accuracy and robustness.

Abstract: As the need for assistance with Activities of Daily Living (ADLs) among older adults increases, smart walkers have emerged as a promising solution. One critical yet underexplored aspect of smart walker design is the handlebar, which is the primary physical interface between the user and the device. This study addresses this gap by introducing an ergonomic handlebar design with biometric sensors and actuators to enhance navigation assistance in smart walkers. The proposed design features a dual-layer handlebar with a hard inner shell for stability and a soft outer shell for user comfort. The left handlebar integrates heart rate and temperature sensors, while the right handlebar incorporates Galvanic Skin Response (GSR) and pressure sensors to monitor physiological responses and physical interaction. Both handlebars are equipped with Force-Sensing Resistors (FSRs) for fall detection and prevention. Two types of haptic feedback—Vibro-Tactile feedback using vibration motors and Skin-Stretch feedback facilitated by a stepper motor and gear-based tractor system—are incorporated to further improve navigation assistance. The system's functionality is validated through sample biometric signals collected from the enhanced handlebars, demonstrating the sensors' effectiveness and the design's user-friendliness.