English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Multi-agent navigation in dynamic environments is of great industrial valuewhen deploying a large scale fleet of robot to real-world applications. Thispaper proposes a decentralized partially observable multi-agent path planningwith evolutionary reinforcement learning (MAPPER) method to learn an effectivelocal planning policy in mixed dynamic environments. Reinforcementlearning-based methods usually suffer performance degradation on long-horizontasks with goal-conditioned sparse rewards, so we decompose the long-rangenavigation task into many easier sub-tasks under the guidance of a globalplanner, which increases agents' performance in large environments. Moreover,most existing multi-agent planning approaches assume either perfect informationof the surrounding environment or homogeneity of nearby dynamic agents, whichmay not hold in practice. Our approach models dynamic obstacles' behavior withan image-based representation and trains a policy in mixed dynamic environmentswithout homogeneity assumption. To ensure multi-agent training stability andperformance, we propose an evolutionary training approach that can be easilyscaled to large and complex environments. Experiments show that MAPPER is ableto achieve higher success rates and more stable performance when exposed to alarge number of non-cooperative dynamic obstacles compared with traditionalreaction-based planner LRA* and the state-of-the-art learning-based method.

MAPPER: Multi-Agent Path Planning with Evolutionary Reinforcement Learning in Mixed Dynamic Environments