• Corpus ID: 247218361

Benchmarking Augmentation Methods for Learning Robust Navigation Agents: the Winning Entry of the 2021 iGibson Challenge

@inproceedings{Yokoyama2021BenchmarkingAM,
  title={Benchmarking Augmentation Methods for Learning Robust Navigation Agents: the Winning Entry of the 2021 iGibson Challenge},
  author={Naoki Yokoyama and Qian Luo and Dhruv Batra and Sehoon Ha},
  year={2021}
}
While impressive progress has been made for teaching embodied agents to navigate static environments using vi-sion, much less progress has been made on more dynamic environments that may include moving pedestrians or mov-able obstacles. In this study, we aim to benchmark different augmentation techniques for improving the agent’s performance in these challenging environments. We show that adding several dynamic obstacles into the scene during training confers significant improvements in test… 

Figures and Tables from this paper

References

SHOWING 1-10 OF 27 REFERENCES

Learning to Explore using Active Neural SLAM

This work presents a modular and hierarchical approach to learn policies for exploring 3D environments, called `Active Neural SLAM'. Our approach leverages the strengths of both classical and

Auxiliary Tasks Speed Up Learning PointGoal Navigation

TLDR
This work develops a method to significantly increase sample and time efficiency in learning PointNav using self-supervised auxiliary tasks (e.g. predicting the action taken between two egocentric observations, predicting the distance between two observations from a trajectory, etc.).

Is Mapping Necessary for Realistic PointGoal Navigation?

TLDR
This paper identifies the main (perhaps, only) cause of the drop in performance: absence of GPS+Compass, and develops human-annotation-free data-augmentation techniques to train models for visual odometry, and scale dataset and model size.

Robot Navigation in Constrained Pedestrian Environments using Reinforcement Learning

TLDR
A compositional principle for multi-layout training is explored and it is found that policies trained in a small set of geometrically simple layouts successfully generalize to more complex unseen layouts that exhibit composition of the structural elements available during training.

Socially aware motion planning with deep reinforcement learning

TLDR
Using deep reinforcement learning, this work develops a time-efficient navigation policy that respects common social norms and is shown to enable fully autonomous navigation of a robotic vehicle moving at human walking speed in an environment with many pedestrians.

Reinforcement Learning with Augmented Data

TLDR
It is shown that augmentations such as random translate, crop, color jitter, patch cutout, random convolutions, and amplitude scale can enable simple RL algorithms to outperform complex state-of-the-art methods across common benchmarks.

Mid-Level Visual Representations Improve Generalization and Sample Efficiency for Learning Visuomotor Policies

TLDR
This work finds that using a mid-level perception confers significant advantages over training end-to-end from scratch (i.e. not leveraging priors) in navigation-oriented tasks and develops an efficient max-coverage feature set that can be adopted in lieu of raw images.

Object Goal Navigation using Goal-Oriented Semantic Exploration

TLDR
A modular system called, `Goal-Oriented Semantic Exploration' which builds an episodic semantic map and uses it to explore the environment efficiently based on the goal object category and outperforms a wide range of baselines including end-to-end learning-based methods as well as modular map- based methods.

On Evaluation of Embodied Navigation Agents

TLDR
The present document summarizes the consensus recommendations of a working group to study empirical methodology in navigation research and discusses different problem statements and the role of generalization, present evaluation measures, and provides standard scenarios that can be used for benchmarking.

Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance?

TLDR
The experiments show that it is possible to tune simulation parameters to improve sim2real predictivity (e.g. improving SRCC from 0.18 to 0.844) – increasing confidence that in-simulation comparisons will translate to deployed systems in reality.