The correspondence problem

By Chrystopher L. Nehaniv, Kerstin Dautenhahn. 2002

300 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Language and Linguistics

View Paper: https://www.semanticscholar.org/paper/656a16142be290c52c14100b8ab673e4060f487e

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Language and Linguistics

Citations in the corpus

AVID: Learning Multi-Stage Tasks via Pixel-Level Translation of Human Videos (2019, methodology): “using explicit pose and object detection [36]–[40], essentially resolving the correspondence problem [41] by instrumenting paired data collection or manually matching hand-specified key points.”
Movement reproduction and obstacle avoidance with dynamic movement primitives and potential fields (2008, background): “The correspondence problem refers to the possible mismatch of link lengths and joints between robot and demonstrator [7].”
A survey of robot learning from demonstration (2009, unknown): “The challenges which arise from these differences are referred to broadly as Correspondence Issues [5].”
Probabilistic model-based imitation learning (2013, background): “One key challenge in imitation learning is the correspondence problem [25]: If the body of the teacher and the robot differ, finding an adequate mapping from the teacher’s demonstrations to the robot is non-trivial.”
Towards a unified science of cultural evolution. (2006, background): “…models appeal to a foundation of mirror-neuron function, leading Whiten (2005a) to question whether they really solve, or instead merely assume to be solved, the essential “correspondence problem” (Nehaniv & Dautenhahn 2002) of mapping between equivalent actions in a model and a cultural learner.”
Recent Advances in Robot Learning from Demonstration (2020, methodology): “This type of learning, often referred to as imitation learning (3, 20), is particularly easy for the demonstrator, requiring almost no training to perform.”
Imitation, empathy, and mirror neurons. (2009, background): “ANRV364-PS60-25 ARI 24 November 2008 18:59 literature inspired by sensory-motor models is the so-called correspondence problem (Nehaniv & Dautenhahn 2002).”
A survey of robot learning from demonstrations for Human-Robot Collaboration (2017, background): “…in the ﬁeld such as the Correspondence Problem that arise due to a mismatch between the teacher’s body conﬁgura-tion and the student’s conﬁguration [4] and interface issue to design user friendly interfaces for demonstration (i.e., motion-capture systems [5]) in order to enable non-robotics…”
One-Shot Imitation from Observing Humans via Domain-Adaptive Meta-Learning (2018, background): “First, the difference in appearance and morphology of the human demonstrator from the robot introduces a systematic domain shift, namely the correspondence problem [29, 6].”
Learning concurrent motor skills in versatile solution spaces (2012, background): “However, simply replaying the demonstration typically does not solve the motor skill task as the robot is unable to exactly reproduce the teacher’s behavior, also known as correspondence problem [24], [25], or because the teacher himself could not optimally solve the task.”