# SEER: Safe Efficient Exploration for Aerial Robots using Learning to Predict Information Gain

**Authors:** Yuezhan Tao, Yuwei Wu, Beiming Li, Fernando Cladera, Alex Zhou, Dinesh Thakur, Vijay Kumar  
**arXiv:** [2209.11034](https://arxiv.org/abs/2209.11034)  
**Date:** 2022-09-22  
**Summary page:** https://www.summarizepaper.com/en/paper/2209.11034v1/

## Key points

- Authors: Yuezhan Tao, Yuwei Wu, Beiming Li, Fernando Cladera, Alex Zhou, Dinesh Thakur, Vijay Kumar
- Challenge: Efficient 3-D exploration in indoor environments for micro aerial vehicles with limited sensing capabilities and payload/power constraints
- Leveraging learning techniques to predict occupancy of unseen areas
- Extracting semantic features
- Sampling viewpoints to anticipate information gains for various exploration objectives
- Devising informative trajectories for safe and intelligent exploration
- SEER framework outperforms existing frameworks by 24% in total path length within a structured indoor environment
- Achieves higher success rate during exploration tasks
- Contribution: Enhancing efficiency and safety of aerial robot exploration in complex indoor settings under resource constraints

## Summary

, , , , 


In their paper titled "SEER: Safe Efficient Exploration for Aerial Robots using Learning to Predict Information Gain," authors Yuezhan Tao, Yuwei Wu, Beiming Li, Fernando Cladera, Alex Zhou, Dinesh Thakur, and Vijay Kumar address the challenge of efficient 3-D exploration in indoor environments for micro aerial vehicles with limited sensing capabilities and payload/power constraints. The researchers introduce an indoor exploration framework that leverages learning techniques to predict the occupancy of unseen areas, extract semantic features, sample viewpoints to anticipate information gains for various exploration objectives, and devise informative trajectories to facilitate safe and intelligent exploration. Through extensive experimentation conducted in both simulated and real-world environments, the proposed approach demonstrates significant performance improvements over existing state-of-the-art exploration frameworks. Specifically, the SEER framework outperforms its counterparts by 24% in terms of total path length within a structured indoor environment and achieves a higher success rate during exploration tasks. This research contributes valuable insights into enhancing the efficiency and safety of aerial robot exploration in complex indoor settings while operating under resource constraints.