AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Differentiable Spatial Planning using Transformers

Devendra Singh Chaplot, Deepak Pathak, Jitendra Malik

ICML 2021 · Architectures and efficiency › RNNs and normalization · approach: Differentiable optimization layers · 5 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use long-range self-attention for spatial value propagation instead of local convolutional propagationSPT versus GPPN · In-distribution overallplanning accuracy: 92.31 -> 99.41Improved
Use long-range self-attention for spatial value propagation instead of local convolutional propagationSPT versus GPPN · Out-of-distribution overallplanning accuracy: 86.95 -> 95.18Improved
Use long-range self-attention for spatial value propagation instead of local convolutional propagationSPT versus GPPN · End-to-end mapping and planning overallplanning accuracy: 63.91 -> 82.29Improved
Train the mapper end-to-end with noisy distances (σ=1) rather than dense, perfect supervisionSPT end-to-end mapping and planning · Manipulationplanning accuracy: 98.42 -> 95.83Hurt
Train the mapper with only 5 trajectories rather than dense supervisionSPT end-to-end mapping and planning · Manipulationplanning accuracy: 98.42 -> 97.18Hurt
Combine noisy distances with supervision from only 5 trajectoriesSPT end-to-end mapping and planning · Manipulationplanning accuracy: 98.42 -> 94.34Hurt
Use sparse supervision from 5 trajectories rather than dense supervisionSPT end-to-end mapping and planning · Manipulationmap accuracy: 98.96 -> 80.15Hurt
Use sparse supervision from 5 trajectories rather than dense supervisionSPT end-to-end mapping and planning · Manipulationplanning accuracy: 98.42 -> 97.18No clear effect
Use SPT rather than DijkstraNavigation planner · M=50 mapsruntime per map (ms): 371.05 -> 18.35Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy