Funded Projects

A large number of swarm robots working under the supervision of a few humans is expected to revolutionize various societal applications ranging from disaster response to pollution clean-up. Human-Swarm Interactions (HSI) however present unique challenges associated with trust attribution, differential mode of interaction (e.g., data vs. natural language), and differential perception of the environment. To resolve these issues, we envision swarm systems with variable autonomy capabilities, which in turn demands mechanisms to sustain adequate levels of trust, and streamline communicative interactions between human partners/supervisors and the swarm agents. The objective of this seed project is to design and develop preliminary components of a closed-loop multi-modal human-swarm interaction platform that allows expressing human intent in a natural yet unambiguous manner, and enable swarm agents to aid in predicting mission-reliability. The specific aims of this seed project are: Aim 1 (lead – Dantu): To identify and evaluate a vocabulary of natural words/gestures for expressing human intent with nominal attention allocation, and a dialog manager that effectively translates the human intent into unambiguous courses-of-action plans for the swarm agents under network link quality constraints. Aim 2 (lead – Chowdhury): To develop algorithms that will allow swarm agents to analyze the risk associated with executing the courses-of-action received from human teammates. Aim 3 (lead - Esfahani): To develop design of human-subject experiments (leveraging virtual environment simulations) for studying the effectiveness of the human intent and the swarm risk analyses models to be created in Aims 1 and 2. The expected outcome is a pool of first-of-its-kind-knowledge on the characteristics of deliberate/adaptive human-swarm-interaction, and a plan to leverage this knowledge for conceiving larger projects that will advance the frontier in the human-swarm interaction domain.