AcTExplore relies solely on tactile sensation to automatically explore and navigate objects when traditional visual cues fade away. We have trained it on primitive objects(cube, sphere), enhancing its proficiency in fundamental movements and enabling exploration across a diverse range of objects. Through sufficient exploration, our algorithm incrementally collects tactile data and reconstructs 3D shapes of the objects as well, which can serve as a representation for higher-level downstream tasks.

While state-of-the-art methods are impressive, i.e., they have no off-road prediction, their generalization to cities outside of the benchmark is unknown. We present a novel method that automatically generates realistic scenes that cause state-of-the-art models go off-road.