Humans have a remarkable ability to fluently engage in joint collision avoidance in crowded navigation tasks despite the complexities and uncertainties inherent in human behavior. Underlying these interactions is a mutual understanding that (i) individuals are prosocial, that is, there is equitable responsibility in avoiding collisions, and (ii) individuals should behave legibly, that is, move in a way that clearly conveys their intent to reduce ambiguity in how they intend to avoid others. The question arises, how can a robot algorithm be developed to demonstrate prosocial and legible characteristics in human-robot interactions. The goal of this research is to develop a novel robot planning algorithm that exhibits these traits, thus allowing it to safely and fluently interact with humans. Specifically, we introduce the notion of a markup factor to incentivize legible and proactive behaviors and an inconvenience budget constraint to ensure equitable collision avoidance responsibility. Our code was written in the Julia Language and integrated with a robot using the Robot Operating System (ROS). Our method is first evaluated in structured simulation environments to evaluate the feasibility of the algorithm. Our approach is then evaluated against well-established multi-agent planning algorithms and it is shown that using our approach produces safe, fluent, and prosocial interactions. We demonstrate the real-time feasibility of our approach with human-in-the-loop simulations (i.e. having humans interact with robots that use the proposed algorithm). Future work involves testing the algorithm in real-world, multi-agent environments. This research into human robot interaction will enable robots to operate more safely around humans in complex environments such as warehouses and hospitals.

With the effects of climate change and increasingly dense populations, large-scale emergencies like natural disasters will displace more human lives through catastrophic building collapse, fires, flooding, and more. In many of these scenarios, search and rescue (SAR) plays a vital role in locating missing persons. This study investigates the use of autonomous Uncrewed Aerial Vehicles (UAVs) equipped with an ultra-wideband (UWB) radar for locating and remotely monitoring the vitals (eg. heart and respiratory rate) of missing persons, as well as the signal processing techniques necessary for robust detection. Initially, we used conventional signal processing methods to extract and analyze respiratory and heart rate data. This approach demonstrates high accuracy (achieving accuracies above 93% in previous experiments), but is limited by the extended stationary hover required to collect data. We expand on this work by using a long short-term memory (LSTM) network that provides an alternative method of vital signs evaluation. The LSTM network classifies persons based on the presence of a heart rate and/or respiration rate. This approach requires less time to collect sufficient sensor data to classify a human's vital sign status while the UAV is on the move. By fusing data from an onboard accelerometer for motion correction, preliminary results indicate a classification accuracy of 89.5% during flight. In future trials, we anticipate that the LSTM-based method will uphold its accuracy levels, even during linear traversal patterns where conventional signal processing methods prove impractical. Furthermore, these techniques pave the way for the future research and development of an effective perception-aware autonomous control algorithm that utilizes ergodic exploration to maximize time spent in areas of interest and uncertainty (e.g., rubble) while optimizing UAV movement for better vital signs detection.