On Learning Optimal Corners in Orthogonal Partially Observable Cooperative Guard Art Galleries
CNN and GATv2 corner heuristics speed CADENCE coverage and cut agent use without losing formal guarantees.
The paper targets unspecified corner choice in CADENCE for the Partially Observable Cooperative Guard Art Gallery Problem. It introduces a CNN that scores grid-encoded candidates and a GATv2 network trained with deep Q-learning on a visibility graph, both preserving coverage and connectivity guarantees. Across 7,500 runs on random orthogonal environments from 50x50 to 250x250, the heuristics beat baseline CADENCE on steps to full coverage and peak agent count, with larger gains at bigger scales. They also improve agent utilization versus Incremental Self-Deployment while supplying formal guarantees ISDA lacks.
- CNN and GATv2-DQN heuristics choose corners while preserving CADENCE guarantees.
- Evaluated across 7,500 runs on maps from 50x50 to 250x250.
- Gains in coverage steps and peak agent count grow with scale.
- Improves agent utilization over ISDA while adding guarantees ISDA lacks.
From the original article
1 min read at arxiv.orgThe CADENCE algorithm solves the Partially Observable Cooperative Guard Art Gallery Problem (POCGAGP) with formal coverage and connectivity guarantees, but leaves unspecified which valid corner each agent should be deployed to, a choice that strongly affects efficiency.
By Yassin Ben Mansour · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.