DSCD-Nav: Dual-Stance Cooperative Debate for Object Navigation

TL;DR

DSCD-Nav tackles overconfident one-shot decision making in zero-shot indoor object navigation under partial observability by introducing a dual-stance cooperative debate between TSU (goal-oriented) and SIB (safety/information-oriented) agents. An NCA arbiter fuses their arguments into a final action, with optional micro-probing for verification when disagreement persists, forming a training-free, interpretable decision loop. The approach is designed as a plug-in atop existing VLM-based pipelines and uses a two-mode execution policy (Mode A and Mode B) to balance progress, safety, and information gain. Across HM3Dv1, HM3Dv2, MP3D, and GOAT, DSCD-Nav delivers consistent improvements in SR and SPL and reduces exploration redundancy (AORI), demonstrating enhanced reliability and efficiency in long-horizon, partial-observation navigation.

Abstract

Adaptive navigation in unfamiliar indoor environments is crucial for household service robots. Despite advances in zero-shot perception and reasoning from vision-language models, existing navigation systems still rely on single-pass scoring at the decision layer, leading to overconfident long-horizon errors and redundant exploration. To tackle these problems, we propose Dual-Stance Cooperative Debate Navigation (DSCD-Nav), a decision mechanism that replaces one-shot scoring with stance-based cross-checking and evidence-aware arbitration to improve action reliability under partial observability. Specifically, given the same observation and candidate action set, we explicitly construct two stances by conditioning the evaluation on diverse and complementary objectives: a Task-Scene Understanding (TSU) stance that prioritizes goal progress from scene-layout cues, and a Safety-Information Balancing (SIB) stance that emphasizes risk and information value. The stances conduct a cooperative debate and make policy by cross-checking their top candidates with cue-grounded arguments. Then, a Navigation Consensus Arbitration (NCA) agent is employed to consolidate both sides' reasons and evidence, optionally triggering lightweight micro-probing to verify uncertain choices, preserving NCA's primary intent while disambiguating. Experiments on HM3Dv1, HM3Dv2, and MP3D demonstrate consistent improvements in success and path efficiency while reducing exploration redundancy.

Figures (10)

• Figure 1: (a) Map-based navigation is brittle to map errors. (b) VLM-based navigation is overly confident, making greedy Top-1 choices. (c) Our DSCD-Nav uses collaborative debate and arbitration for more reliable decisions.
• Figure 2: DSCD-Nav overview.(I) Candidate and context construction packages the goal, optional memory, and evidence cues, and a geometry-pruned set of executable polar action candidates with language descriptions into debate-ready inputs. (II) TSU and SIB conduct multi-round cooperative debate over the shared candidates, exchanging preferences and cue-grounded evidence while updating beliefs. (III) The NCA judge arbitrates the debate to produce a final action with a concise rationale and supporting evidence, after which the system executes it directly under consensus or applies bounded geometric soft-compromise with micro-probing when disagreement persists, and the angular gap is within a preset threshold.
• Figure 3: Qualitative case study on HM3Dv2 ObjectNav. We overlay the executed trajectory on the reconstructed 3D scene and annotate several decision points with the corresponding TSU and SIB preferences and the NCA arbitration result. The trace highlights DSCD-Nav’s ability to resolve disagreement at visually ambiguous viewpoints and, when needed, trigger a conservative micro-probing step before committing to the executed action.
• Figure 4: An 8-step, debate-guided HM3Dv1 ObjectNav episode from the agent’s first-person view, illustrating sequential observations, executed actions, and successful termination at the target.
• Figure 5: Top-down trajectory visualization. The floor-plan view overlays the agent’s navigation trace within the scene. Colored arrows denote the executed motions and a small set of salient alternative directions at key junctions, while red crosses mark branches that are discarded or lead to failures or inefficient detours.