arXiv:2603. 18624v2 Announce Type: replace-cross Abstract: Zero-shot object-goal navigation (ZSON) requires navigating unknown environments to find a target object without task-specific training.
By Shuqi Xiao, Maani Ghaffari, Chengzhong Xu, Hui Kong
The paper introduces a unified branch‑and‑bound framework for the Steiner Traveling Salesman Problem on Graphs of Convex Sets (GCS), where the goal is to find a minimum‑cost closed walk through required convex sets while allowing optional vertices and revisits. The method uses additive lower‑bound graph costs for committed prefixes and a cut‑separated connected‑flow relaxation for the remaining cost, guaranteeing finite termination under a uniform positive‑cost assumption. Experiments on benchmark instances show that both best‑first and depth‑first traversal strategies find feasible solutions within 30 seconds, achieving mean certified optimality gaps of 28.1% and 29.7% respectively, outperforming two recent baselines.
By Jingtao Tang, Hang Ma
arXiv:2608. 06702v1 Announce Type: cross Abstract: Lifelong Multi-Agent Path Finding (LMAPF) requires generating collision-free paths for large agent fleets under strict real-time constraints.
By Vaibhav Sanjay, Jiaoyang Li
The paper introduces DCRL (Divide-and-Conquer RL), a method that recursively decomposes offline goal-conditioned reinforcement learning trajectories into a balanced binary tree. By training values from the leaves up to the root, DCRL avoids noisy max-based backups and reduces bootstrap depth from linear to logarithmic, thereby limiting error accumulation. Experiments on diverse goal-reaching tasks show that DCRL outperforms prior flat offline GCRL methods, achieving a higher average score on the most challenging long-horizon OGBench tasks.
By Hyeonseong Jeon, Youngwoon Lee
The paper presents DORA, an online learning algorithm for robot navigation that uses Dijkstra’s algorithm as an exact planning engine under a weaker condition than usual causality—specifically, nonnegativity of a reduced cost on a determinized map. DORA calls a shortest‑path oracle a fixed number of times per episode, avoids estimating transition kernels, and incorporates a logarithmic survival weight to keep contact probabilities with dynamic obstacles within a budget. Experiments on grid‑world, directional drilling, and drone surveillance benchmarks show that DORA matches optimistic value iteration with the true transition kernel while performing 4.5 to 19.3 times less planner work, reduces contacts by a factor of seventeen compared to determinize‑and‑replan, and maintains contact rates within wide budget ranges.
By Mansur M. Arief, Ali Akarma, Ahmad Alfan Alfian Irfan
HorizonFlow is a hierarchical planner for offline goal-conditioned reinforcement learning that treats the planning horizon as an output rather than a fixed input. It uses a subgoal route planner and an action-prefix controller, both employing insertion-based generation and flow matching, to jointly generate continuous plan content and its length. The method leverages the partially generated plan to guide token insertion and to steer generation toward shorter plans, achieving superior performance on Maze2D, Multi2D, and OGBench benchmarks.
By JunHyeok Oh, Zian Jang, Byung-Jun Lee