Human motion is by so far unpredictable to our eyes, how if an autonomous agent is introduced in a crowd and is tasked to navigate through.
Learning the causality of human motion goes beyond observations. One need to collect interventions on human beings and learn a causal model from the observed reactions. Until then, an autonomous system having had built a causal model of its environment will be able to reason better on the state (or predictable states) of its environment, and thereafter make optimal decisions on its actions. In order to build that causal model, the student will work with the real Tiago Robot from PAL Robotics Ltd. https://pal-robotics.com/ by designing real scenarios to let a human being intervene on the environment observed by Tiago. Data collection from interventions will be used by the student to extract initially humans motions from RGB images or Lidar by using state-of-art skeleton detection networks.
Interested students can contact Luca Castri @ lcastri@lincoln.ac.uk and/or myself @ smghames@lincoln.ac.uk