Mobile robots are increasingly deployed in human-centric indoor environments such as homes, hospitals, and labs, where success depends not only on geometric feasibility but on maintaining sufficient clearance while executing a trajectory. Crucially, the required clearance is context dependent: local clutter, maneuvering demands (e.g., sharp turns in narrow passages), and momentary perception reliability can make the same measured distance safe in one situation and risky in another. This is amplified in vision-based navigation, where depth and egomotion estimates vary with viewpoint, motion, and sensing artifacts, so minimum measured clearance is a noisy proxy for executability. As a result, hard selection rules with a globally fixed clearance threshold oscillate between two failure modes: overly conservative behavior in open areas (detours and timeouts) and brittle behavior in dense clutter (risky shortcuts under perception bias).
We evaluate PointGoal navigation on HM3D and MP3D, in-domain and under cross-dataset transfer, where our method achieves the highest SR and SPL among strong diffusion, optimization-based, and reinforcement-learning baselines. We further demonstrate sim-to-real transfer on a Unitree G1 humanoid in cluttered indoor scenes without task-specific tuning, and our ablations isolate the contribution of each critic component.
End-to-End Visual Navigation
End-to-end visual navigation closes the perception–action loop by mapping egocentric observations to low-level actions or short-horizon motion segments, reducing latency and error propagation across perception, mapping, and planning. Large-scale simulators such as Habitat provide standardized embodied tasks, where deep RL agents (e.g., DD-PPO) set strong PointGoal baselines, and hybrid systems such as Active Neural SLAM add explicit spatial memory via learned mapping and hierarchical policies over a top-down map.
Generate–Select Framework
We introduced a generate–select framework that replaces fixed margins with a learnable, context-conditioned safety critic fusing safety and efficiency in balance. Across simulation and real-world tests it achieves the highest SR and SPL, and transfers from pure simulation to a real humanoid in cluttered indoor scenes without task-specific tuning. Ablations confirm each critic term contributes to a favorable safety–efficiency trade-off and consistent gains over fixed-threshold baselines.
Sim-to-Real Transfer on Unitree G1
We demonstrate successful sim-to-real transfer on a Unitree G1 humanoid in cluttered indoor scenes without any task-specific tuning. The learned safety margins adapt to the cluttered environment, enabling the humanoid to maintain appropriate clearance while navigating through tight spaces and around obstacles.
Frequently Asked Questions
How does adaptive safety margin selection differ from fixed-threshold approaches? Adaptive safety margins use a learned, context-conditioned critic that adjusts clearance requirements based on local clutter, maneuvering demands, and perception reliability, whereas fixed-threshold methods apply a single global distance threshold regardless of situation.
What benchmark datasets were used to evaluate the method? The method was evaluated on PointGoal navigation tasks using the HM3D and MP3D datasets, including both in-domain and cross-dataset transfer scenarios.
Can the method transfer directly from simulation to real robots without retuning? Yes, the paper demonstrates successful sim-to-real transfer on a Unitree G1 humanoid in cluttered indoor scenes without any task-specific tuning.
What components contribute to the safety-efficiency trade-off in the framework? The ablation studies confirm that each critic term independently contributes to achieving a favorable balance between safety and efficiency, with consistent improvements over fixed-threshold baselines.