M.Sc. Thesis: Behavior Trees for Assisted Robot Teleoperation

People with tetraplegia or other physical impairment who lack limb controlneed assistance in order to executes everyday tasks and Actions of DailyLiving (ADLs). Unfortunately, having human assistance is not reliable dueto the human assistant or care giver’s availability throughout the day. Ad-ditionally, it could be quite expensive since experienced care givers are ingreat demand. Research recently started leaning towards robotic assistance for the physically impaired. This comes in the form of robot arm teleop-eration with varying levels of autonomy interleaved with the direct human control.

Adding new actions to these frameworks, however, requires extensive robot programming knowledge in order to develop the motion primitives. In this thesis you will create an extension of Behavior Trees (BTs) that offers assistance during robot teleoperation. BT is a modular graphical robot policy representation. This offers an easy way to model new actions with little programming knowledge.

Your Task:

The goal of this thesis is to extend BTs using Teleoperation Action Nodes(TAN) that offers assistance to the operator in terms of modifying theirinput signal. Part of your work is to implement a user interface allowingthe visualization of existing BTs as well as creating new ones. You will implement and evaluate this extension on a Kinova Jaco robot arm. For example, the action of pouring drinks from a bottle, where the bottle needs to be rotate on different pivot points before the liquid can be poured -safely- and then the bottle is again rotated around its base back to its upright position can be represented using a BT that sequences the applications of the rotate action using different parameters

Requirements:

  • Willingness to learn
  • Solid Pyhton knowledge
  • Knowledge of machine learning
  • Experience with a machine learning framework is a plus
  • Knowledge of ROS and experience with one or more robot platforms are a plus

Further Information

If you are interested in it or have questions, please feel free to contact Mohamed Behery or Minh Trinh (WZL) via email: