Geri Dön

Adaptive force trajectory tracking for robotic interaction with soft objects

Başlık çevirisi mevcut değil.

PDF İndir

  1. Tez No: 785839
  2. Yazar: NOYAN ÜZER
  3. Danışmanlar: DR. YİANNİS KARAYİANNİDİS
  4. Tez Türü: Yüksek Lisans
  5. Konular: Elektrik ve Elektronik Mühendisliği, Electrical and Electronics Engineering
  6. Anahtar Kelimeler: Adaptive force control, neuro-adaptive force control, robot force control, velocity based force control, neural network based function approximation, contact modeling
  7. Yıl: 2021
  8. Dil: İngilizce
  9. Üniversite: Chalmers Unıversıty Of Technology
  10. Enstitü: Yurtdışı Enstitü
  11. Ana Bilim Dalı: Elektrik Mühendisliği Ana Bilim Dalı
  12. Bilim Dalı: Belirtilmemiş.
  13. Sayfa Sayısı: Belirtilmemiş.

Özet

Özet yok.

Özet (Çeviri)

Controlling the interaction between a robotic manipulator and its environment is a necessity for the execution of a plethora of practical tasks. However, without the knowledge of the contact dynamics, sufficient control performance is hard to achieve. A pure motion control strategy can perhaps be favored in this case. If, however, the task is not carefully planned, this may end up with the breakage of the robot. A more preferable approach to take would be to use a mathematical model to describe the dynamics of the environment then based on this model design a control law which is capable of adaptively adjusting the position/velocity trajectory followed by the end-effector depending on the contact force readings. In this thesis, the latter strategy was followed. In order to approximate the environment dynamics, two models were used: The linear Kelvin-Voight model and the nonlinear Hunt-Crossley model. The linear model while being widely used and easy to work with was shown not to be as consistent with physics of contact as its nonlinear counterpart. Based on these contact models two sets of force controllers were proposed for velocity/position controlled robotic manipulators: A set of KelvinVoigt based adaptive controllers and a set of Hunt-Crossley based neuro-adaptive controllers. It was shown that when the robot is under exact or asymptotically exact inner loop velocity control, the controllers provide asymptotic force trajectory tracking. Finally, a simple artificial neural network was designed to approximate the dynamical properties of a compliant environment, in this case a sponge, for the purpose of creating a testing environment. A series of comparative experiments were performed and the step responses and tracking capabilities of the controllers were investigated.

Geri Dön