
Join us as we wrap up our season with an engaging paper that connects themes from our mini-track: How does a robot determine where to interact based on sensed contact and force?
Our featured paper, Coupled Particle Filters for Robust Affordance Estimation (Lowin, Mengers & Brock, 2026), is produced by Oliver Brock's Robotics and Biology Laboratory at TU Berlin and will be presented at ICRA 2026.
✨ Key Insights: Affordance estimation — figuring out where a robot can grasp and manipulate an object — is a complex challenge due to the ambiguity in visual input. The authors propose a clever solution using two coupled particle filters: one for identifying "graspable" areas and the other for "movable" regions. These filters exchange information bidirectionally, honing in on areas of agreement where affordances exist.
🔍 Results: Evaluated against a real-world dataset, this innovative method significantly outperforms three recent approaches (Where2Act, Hands-as-Probes, and HRP) under challenging conditions, including low lighting.
As we conclude our season, let's reflect on how these discussing topics—sensing contact acoustically, developing force-sensitive manipulation strategies, and understanding affordances—interconnect within a complete manipulation pipeline.
💭 Questions to Ponder: Is coupling simpler, interpretable estimators more effective than complex learned systems? How should we balance physical priors with data-driven approaches?
Join us for a lively discussion! We're looking forward to your thoughts!
📍 Location: Merantix AI Campus
🎟️ Get your ticket here!
Merantix AI Campus
TrasaMerantix AI Campus
Trasa