TRACE: Ergodic Trajectory Optimization for Active Scene Reconstruction

2026-08-03

AI Summary

This work addresses the inefficiency of existing active Gaussian splatting reconstruction methods, which rely on greedy viewpoint selection and neglect global scene structure, often leading to suboptimal trajectories. To overcome this limitation, the paper formulates active reconstruction as an ergodic coverage problem and introduces a kernel-based ergodic planner that optimizes sensor trajectories such that their time-averaged spatial distribution matches a target information density—dynamically generated online from the uncertainty and visibility of the current Gaussian splatting map. By integrating ergodic control into active scene reconstruction for the first time, the approach establishes a closed-loop coupling between perception and planning, thereby avoiding local optima. Combined with gradient-flow optimization and a footprint exhaustion mechanism, the method achieves a 1.5 dB PSNR improvement over the NBV baseline on the Replica dataset, significantly enhancing both reconstruction quality and perceptual efficiency.

Abstract

Existing active reconstruction systems with Gaussian-splatting maps select observations greedily, optimizing a single next-best-view (NBV) at each step and connecting the chosen views by short-horizon path planning. This greedy decoupling disregards the global structure of scene information, producing inefficient trajectories that waste sensing capacity in transit between selected views. In this work, we study active reconstruction as an ergodic coverage problem: the time-averaged spatial statistics of the sensor trajectory should match a target information distribution induced by the current map. Our approach derives this target distribution online from uncertainty and visibility, and calculates ergodic trajectories via a kernel-ergodic horizon planner with gradient flow and footprint depletion, closing the loop between mapping and trajectory optimization. We thoroughly evaluate TRACE on the Replica dataset against the Next-Best-View (NBV) baselines, improving PSNR by 1.5 dB. Code: https://github.com/spikelab-jhu/trace-active-reconstruction.

Problem

• active reconstruction
• next-best-view
• ergodic coverage
• trajectory optimization
• scene information

Innovation

• ergodic trajectory optimization
• active scene reconstruction
• Gaussian splatting
• next-best-view
• information-driven planning