Mapping Navigation Instructions to Continuous Control Actions with Position-Visitation Prediction
This addresses the challenge of precise drone navigation from instructions, though it appears incremental as it builds on existing instruction-following methods.
The paper tackles the problem of mapping natural language instructions to continuous control actions for quadcopter drones by predicting interpretable position-visitation distributions, achieving a 16.85% absolute improvement in task-completion accuracy over state-of-the-art methods in simulation.
We propose an approach for mapping natural language instructions and raw observations to continuous control of a quadcopter drone. Our model predicts interpretable position-visitation distributions indicating where the agent should go during execution and where it should stop, and uses the predicted distributions to select the actions to execute. This two-step model decomposition allows for simple and efficient training using a combination of supervised learning and imitation learning. We evaluate our approach with a realistic drone simulator, and demonstrate absolute task-completion accuracy improvements of 16.85% over two state-of-the-art instruction-following methods.