EvoNav-Bench extends GOAT-Bench style lifelong navigation to environments that change between tasks, built on ProcTHOR, so earlier experience is useful but no longer fully reliable. It benchmarks three recent methods that build and reuse scene representations, and three simple strategies for handling change: Frontier-Update, Fail-then-Update and Stage-Reset. Existing methods turn out brittle when the environment evolves.
Relevance
Lifelong navigation methods typically assume a stationary world, and this benchmark measures what happens when that assumption fails. For long-term autonomy and embodied AI benchmarks.
Links
- arXiv
- Subfield: Planning and navigation, Semantic mapping, Robot learning