analiza aiqbench
This research challenges the standard deep reinforcement learning practice of immediately resetting an environment upon collision, proposing a budgeted approach that allows agents to retry difficult configurations within a single episode. By decoupling local failures from global resets, the proposed Multi-Collision reset Budget framework significantly improves early-stage learning efficiency and reduces the total interactions needed to achieve target success rates. The method demonstrates practical viability across both simulations and heterogeneous real-world robot platforms, offering a more robust training strategy for navigating cluttered environments.
Do We Really Need Immediate Resets? Rethinking Collision Handling for Efficient Robot Navigation
arXiv:2605.02192v2 Announce Type: replace Abstract: Should a single collision necessarily terminate an entire navigation episode? In most deep reinforcement learning (DRL) frameworks for robot navigation, this remains the standard practice: every collision immediately triggers a global environment reset and is penalized as a complete task failure. While a collision during deployment naturally indicates task failure, applying the same treatment during training prevents the agent from exploring challenging obstacle configurations, which slows learning progress in the early training phase. In this work, we challenge this convention and propose a Multi-Collision reset Budget (MCB) framework that decouples local collision termination from global environment resets, allowing the agent to retry difficult configurations within the same episode. Simulation experiments show that MCB improves early-stage learning efficiency by reaching target success-rate levels with fewer interactions, with small collision budgets producing the most consistent gains. Real-world experiments on heterogeneous robot platforms further validate the deployability of the learned policies in cluttered environments.
Title:Do We Really Need Immediate Resets? Rethinking Collision Handling for Efficient Robot Navigation
Abstract:Should a single collision necessarily terminate an entire navigation episode? In most deep reinforcement learning (DRL) frameworks for robot navigation, this remains the standard practice: every collision immediately triggers a global environment reset and is penalized as a complete task failure. While a collision during deployment naturally indicates task failure, applying the same treatment during training prevents the agent from exploring challenging obstacle configurations, which slows learning progress in the early training phase. In this work, we challenge this convention and propose a Multi-Collision reset Budget (MCB) framework that decouples local collision termination from global environment resets, allowing the agent to retry difficult configurations within the same episode. Simulation experiments show that MCB improves early-stage learning efficiency by reaching target success-rate levels with fewer interactions, with small collision budgets producing the most consistent gains. Real-world experiments on heterogeneous robot platforms further validate the deployability of the learned policies in cluttered environments.
Robotics (cs.RO)
arXiv:2605.02192 [cs.RO]
(or arXiv:2605.02192v2 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2605.02192 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 255 cuvinte
Subiecte conexe
Rethinking Embodied Navigation via Relational Inductive Bias
arXiv · 10 iun. 2026
451 Research report: How agentic AI is redefining identity security
1Password · 20 aug. 2026
The Builders Stage brings practical strategies for scaling startups to TechCrunch Disrupt 2026
TechCrunch · 2 sept. 2026
Rethinking Legibility in Social Robot Hallway Navigation: Impact of Intent Representation and Human Distraction
arXiv · 1 oct. 2026