Artificial Intelligence and Human Interaction

Combining supervised and reinforcement learning with human interaction modalities such as task demonstrations, eye gaze data, and natural language, to safely train autonomous systems.

Human-in-the-loop Machine Learning

Cycle-of-Learning for Autonomous Systems from Human Interaction: a concept for combining multiple forms of human interaction with reinforcement learning. As the policy develops, the autonomy independence increases and the human interaction level decreases.

Performance comparison in terms of task completion and samples required with Interventions (Int), Demonstrations (Demo) and the Cycle-of-Learning (CoL) framework for (A) 4 human interactions, (B) 8 human interactions, (C) 12 human inter-actions and (D) 20 human interactions, respectively.

Comparison of CoL, BC, DDPG, and DAPG for 3 random seeds in the dense- and sparse-reward LunarLanderContinuous-v2 environment, respectively.

Integrating Behavior Cloning and Reinforcement Learning for Improved Performance in Sparse Reward Environments

Efficiently Combining Human Demonstrations and Interventions for Safe Training of Autonomous Systems in Real-Time

Cycle-of-Learning for Autonomous Systems from Human Interaction

Cyber-Human Approach For Learning Human Intention And Shape Robotic Behavior Based On Task Demonstration

PODNet: A Neural Network for Discovery of Plannable Options**

Intelligent Motion Video Guidance for Unmanned Air System Ground Target Tracking

Target Tracking with Reinforcement Learning

Reinforcement learning agent trained to control a real fixed-wing aircraft and perform video tracking of ground targets.

Morphing Wing

More information soon.