Contact Driven Robotic Manipulation

Friday, 1 April, 2016

Developing a robot that has close-to-human manipulation abilities would change the way robots are used in the current society. The potential of such machines could transform the life of most of the developed countries inhabitants. Factories could include such robots into their manufacturing lines, move the workers performing repetitive tasks to a higher level duty, and increase their productivity. In addition, robots could start being deployed in real houses to assist elderly or disabled people boosting their quality of life and giving them back some independence. Maybe in the future, service robots will be present in each house and considered another house appliance like the washing machine or the dishwasher.

In this thesis we have implemented and validated a contact driven robotic manipulation system. First, in Chapter 2 we have taken inspiration from neuroscience studies about human sensorimotor control of manipulation and identified the key components of a contact driven manipulation system. Second, in Chapter 3 we have presented, implemented and validated the manipulation primitives paradigm, a vocabulary of simple sensor-based manipulation actions that are combined to perform complex tasks. Third, we have developed and validated the mechanisms for contact event detection and prediction in Chapters 4 and 5 respectively. Fourth, in Chapter 6 we have presented the software architecture created to integrate all the pieces of the system and some useful abilities such as grasp planning. Fifth, in Chapter 7 an abstraction mechanism that allows the same tasks to be used by different robotic platforms has been presented. Finally, to endow the contact-based manipulation system with visual perception of objects we have presented in Chapter 8 a hierarchical object recognition system based on primate brain mechanisms. Details about the robotic platforms used to develop the work presented in this thesis are provided in Appendix A.
for more details here