Learning Your Way Without Map or Compass: Panoramic Target Driven Visual Navigation

Watkins-Valls, David; Xu, Jingxi; Waytowich, Nicholas R.; Allen, Peter K.

doi:10.1109/iros45743.2020.9341511

Cited by 14 publications

(9 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The community has built several indoor navigation simulators [41,57,40,27] on top of photo-realistic scans of 3D environments [27,6,47,56,55]. To test a robot's ability to perceive, navigate and interact with the environment, the community has also introduced several tasks [57,5,45,10,52,36,3,28,48,22,21,51,16,34,33,31,32] and benchmarks. Specifically, Batra et al [5] introduce evaluation details for the task of Object Navigation, requiring the agent to navigate to a given object class instead of a final point-goal.…”

Section: Related Workmentioning

confidence: 99%

The Surprising Effectiveness of Visual Odometry Techniques for Embodied PointGoal Navigation

Zhao¹,

Agrawal

Batra

et al. 2021

Preprint

View full text Add to dashboard Cite

It is fundamental for personal robots to reliably navigate to a specified goal. To study this task, PointGoal navigation has been introduced in simulated Embodied AI environments. Recent advances solve this PointGoal navigation task with near-perfect accuracy (99.6% success) in photo-realistically simulated environments, assuming noiseless egocentric vision, noiseless actuation and most importantly, perfect localization. However, under realistic noise models for visual sensors and actuation, and without access to a "GPS and Compass sensor," the 99.6%-success agents for PointGoal navigation only succeed with 0.3%. 1 In this work, we demonstrate the surprising effectiveness of visual odometry for the task of PointGoal navigation in this realistic setting, i.e., with realistic noise models for perception and actuation and without access to GPS and Compass sensors. We show that integrating visual odometry techniques into navigation policies improves the state-of-the-art on the popular Habitat PointNav benchmark by a large margin, improving success from 64.5% to 71.7% while executing 6.4 times faster.

show abstract

Section: Related Workmentioning

confidence: 99%

The Surprising Effectiveness of Visual Odometry Techniques for Embodied PointGoal Navigation

Zhao¹,

Agrawal

Batra

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…To train PointNav policy that does not require a semantic sensor, we used the faster version of Habitat (that does not support native semantic segmentation sensor), the BPS simulator (Shacklett et al 2021), which was 100x timed faster. With a similar approach, these works (Kadian et al 2020) (Watkins-Valls et al 2020 showed the ability to transfer the RL model from a simulated environment to real-world usage.…”

Section: Related Workmentioning

confidence: 99%

Landmark Policy Optimization for Object Navigation Task

Staroverov,

Panov

2021

Preprint

View full text Add to dashboard Cite

This work studies object goal navigation task, which involves navigating to the closest object related to the given semantic category in unseen environments. Recent works have shown significant achievements both in the end-to-end Reinforcement Learning approach and modular systems, but need a big step forward to be robust and optimal. We propose a hierarchical method that incorporates standard task formulation and additional area knowledge as landmarks, with a way to extract these landmarks. In a hierarchy, a low level consists of separately trained algorithms to the most intuitive skills, and a high level decides which skill is needed at this moment. With all proposed solutions, we achieve a 0.75 success rate in a realistic Habitat simulator. After a small stage of additional model training in a reconstructed virtual area at a simulator, we successfully confirmed our results in a real-world case.

show abstract

“…A fouth action called "Done" is executed whenever the agent is within 0.2 m from the goal position. Note that in [1], they trained an agent to also learn this trivial task, but this was not the case in [11], [6]; we take the latter approach. A schematic of the overall setup is shown in Fig.…”

Section: Problem Setupmentioning

confidence: 99%

“…They demonstrated robust learning of both perception and policy on all three tasks, including transfer to new visual environments as well as to new embodied tasks. In [11], the authors undertook Imitation Learning to train a robot to navigate in the Gibson simulator using the Dijkstra algorithm and obtained high success rates. In another study, the authors used DRL for target-driven robot navigation in an indoor scenes simulator called AI2THOR [12], where only RGB images of the state and the target are used to train the navigation policy network.…”

Section: Introductionmentioning

confidence: 99%

An A* Curriculum Approach to Reinforcement Learning for RGBD Indoor Robot Navigation

Balakrishnan,

Chakravarty,

Shrivastava

2021

Preprint

View full text Add to dashboard Cite

Training robots to navigate diverse environments is a challenging problem as it involves the confluence of several different perception tasks such as mapping and localization, followed by optimal path-planning and control. Recently released photo-realistic simulators such as Habitat [1] allow for the training of networks that output control actions directly from perception: agents use Deep Reinforcement Learning (DRL) to regress directly from the camera image to a control output in an end-to-end fashion. This is data-inefficient and can take several days to train on a GPU. Our paper tries to overcome this problem by separating the training of the perception and control neural nets and increasing the path complexity gradually using a curriculum approach. Specifically, a pre-trained twin Variational AutoEncoder (VAE) [2] is used to compress RGBD (RGB & depth) sensing from an environment into a latent embedding, which is then used to train a DRLbased control policy. A*, a traditional path-planner is used as a guide for the policy and the distance between start and target locations is incrementally increased along the A* route, as training progresses. We demonstrate the efficacy of the proposed approach, both in terms of increased performance and decreased training times for the PointNav task in the Habitat simulation environment. This strategy of improving the training of direct-perception based DRL navigation policies is expected to hasten the deployment of robots of particular interest to industry such as co-bots on the factory floor and last-mile delivery robots. 2

show abstract

Learning Your Way Without Map or Compass: Panoramic Target Driven Visual Navigation

Cited by 14 publications

References 20 publications

The Surprising Effectiveness of Visual Odometry Techniques for Embodied PointGoal Navigation

The Surprising Effectiveness of Visual Odometry Techniques for Embodied PointGoal Navigation

Landmark Policy Optimization for Object Navigation Task

An A* Curriculum Approach to Reinforcement Learning for RGBD Indoor Robot Navigation

Contact Info

Product

Resources

About