Unsupervised Hump Detection for Mobile Robots Based On Kinematic Measurements and Deep-Learning Based Autoencoder

Rettig, Oliver; Müller, Silvan; Strand, Marcus; Katić, Darko

doi:10.1007/978-3-030-01370-7_9

Cited by 1 publication

(4 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The bigger aim of the project behind this paper is to make the usage of mobile robots more robust and flexible by dynamic adaptions to a changing environment. This paper extends the work in [13], which describes in detail the kinematics of the commercially available mobile platform Mir-100 during overrun of a cable channel as a model for an environmental anomaly. Takeoffs are happening particular strong for the rear wheels as a product of the front and the drive wheels already past the cable channel and therefore pulling is more effectively.…”

Section: Conceptmentioning

confidence: 62%

“…In [13] we have shown that a specific deep neural network (DNN) based autoencoder allow for a robust and easily expandable implementation of anomaly detection in kinematic data but which architecture should we use?…”

Section: Methodsmentioning

confidence: 99%

“…1) in a gait-and motion analysis lab. Details of the dataset and its aquisition is described in [13]. Three trials are arbitrary chosen to build a validation set.…”

Section: Methodsmentioning

confidence: 99%

“…The measurements described in [13] are done with high precision by a marker based optical system to have a "gold standard". This dataset is also used to train the DNNs presented in this paper.…”

Section: Conceptmentioning

confidence: 99%

See 3 more Smart Citations

Which deep artifical neural network architecture to use for anomaly detection in Mobile Robots kinematic data?

Rettig

Müller

Strand

et al. 2018

Machine Learning for Cyber Physical Systems

Self Cite

View full text Add to dashboard Cite

Small humps on the floor go beyond the detectable scope of laser scanners and are therefore not integrated into SLAM based maps of mobile robots. However, even such small irregularities can have a tremendous effect on the robot's stability and the path quality. As a basis to develop anomaly detection algorithms, kinematics data is collected exemplarily for an overrun of a cable channel and a bulb plate. A recurrent neuronal network (RNN), based on the autoencoder principle, could be trained successfully with this data. The described RNN architecture looks promising to be used for realtime anomaly detection and also to quantify path quality.

show abstract

Section: Conceptmentioning

confidence: 62%

Section: Methodsmentioning

confidence: 99%

“…1) in a gait-and motion analysis lab. Details of the dataset and its aquisition is described in [13]. Three trials are arbitrary chosen to build a validation set.…”

Section: Methodsmentioning

confidence: 99%

Section: Conceptmentioning

confidence: 99%

See 2 more Smart Citations