Search papers, labs, and topics across Lattice.
The paper introduces ESCAPE, a novel architecture for long-horizon mobile manipulation that combines episodic spatial memory with an adaptive execution policy. ESCAPE constructs a persistent, depth-free 3D spatial memory using a Spatio-Temporal Fusion Mapping module and grounds targets using a Memory-Driven Target Grounding module. The adaptive policy orchestrates global navigation and local manipulation, achieving state-of-the-art performance on the ALFRED benchmark, particularly in long-horizon tasks with limited guidance.
Forget rigid policies: ESCAPE's adaptive execution and spatial memory enable SOTA long-horizon mobile manipulation, even without detailed instructions.
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.
