ESCAPE: Episodic Spatial Memory and Adaptive Execution Policy for Long-Horizon Mobile Manipulation
This work improves long-horizon task performance for embodied AI agents in complex indoor environments, but the gains are incremental over existing methods.
ESCAPE addresses catastrophic forgetting, spatial inconsistency, and rigid execution in long-horizon mobile manipulation, achieving state-of-the-art success rates of 65.09% and 60.79% on the ALFRED benchmark in test seen and unseen environments, respectively.
Coordinating navigation and manipulation with robust performance is essential for embodied AI in complex indoor environments. However, as tasks extend over long horizons, existing methods often struggle due to catastrophic forgetting, spatial inconsistency, and rigid execution. To address these issues, we propose ESCAPE (Episodic Spatial Memory Coupled with an Adaptive Policy for Execution), operating through a tightly coupled perception-grounding-execution workflow. For robust perception, ESCAPE features a Spatio-Temporal Fusion Mapping module to autoregressively construct a depth-free, persistent 3D spatial memory, alongside a Memory-Driven Target Grounding module for precise interaction mask generation. To achieve flexible action, our Adaptive Execution Policy dynamically orchestrates proactive global navigation and reactive local manipulation to seize opportunistic targets. ESCAPE achieves state-of-the-art performance on the ALFRED benchmark, reaching 65.09% and 60.79% success rates in test seen and unseen environments with step-by-step instructions. By reducing redundant exploration, our ESCAPE attains substantial improvements in path-length-weighted metrics and maintains robust performance (61.24% / 56.04%) even without detailed guidance for long-horizon tasks.