# How to Train PointGoal Navigation Agents on a (Sample and Compute) Budget

arXiv: https://arxiv.org/abs/2012.06117 · HTML: https://mlsanity.com/paper/2012.06117/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 11 Dec 2020.

## Built on by (most cited first, 7 of 7)

- [Habitat 2.0: Training Home Assistants to Rearrange their Habitat](https://mlsanity.com/paper/2106.14405/) (2021)
- [Simple but Effective: CLIP Embeddings for Embodied AI](https://mlsanity.com/paper/2111.09888/) (2021)
- [Auxiliary Tasks and Exploration Enable ObjectNav](https://mlsanity.com/paper/2104.04112/) (2021)
- [Retrospectives on the Embodied AI Workshop](https://mlsanity.com/paper/2210.06849/) (2022)
- [Is Mapping Necessary for Realistic PointGoal Navigation?](https://mlsanity.com/paper/2206.00997/) (2022)
- [Towards Disturbance-Free Visual Mobile Manipulation](https://mlsanity.com/paper/2112.12612/) (2021)
- [Landmark Policy Optimization for Object Navigation Task](https://mlsanity.com/paper/2109.09512/) (2021)

## Abstract (arXiv)

PointGoal navigation has seen significant recent interest and progress, spurred on by the Habitat platform and associated challenge. In this paper, we study PointGoal navigation under both a sample budget (75 million frames) and a compute budget (1 GPU for 1 day). We conduct an extensive set of experiments, cumulatively totaling over 50,000 GPU-hours, that let us identify and discuss a number of ostensibly minor but significant design choices -- the advantage estimation procedure (a key component in training), visual encoder architecture, and a seemingly minor hyper-parameter change. Overall, these design choices to lead considerable and consistent improvements over the baselines present in Savva et al. Under a sample budget, performance for RGB-D agents improves 8 SPL on Gibson (14% relative improvement) and 20 SPL on Matterport3D (38% relative improvement). Under a compute budget, performance for RGB-D agents improves by 19 SPL on Gibson (32% relative improvement) and 35 SPL on Matterport3D (220% relative improvement). We hope our findings and recommendations will make serve to make the community's experiments more efficient.