Mobile Manipulation Hackathon (IROS 2018) - AUDECO-IOC-UPC Team

In this talk

1. 1:09

   Choosing a table approach

   The team says the robot plans approaches from four sides of the table and selects one based on its capabilities. They describe using a robustness index to choose a safer side, though they skip that selection during the demonstration and attach manually.

2. 2:02

   Grasping a selected drink

   A person selects a drink by keyboard, and the team demonstrates the robot reaching for the requested item. They describe a yellow gripper with a label and force sensor, intended to improve GIoTTO’s manipulation and grasping.

3. 4:09

   A serving attempt fails

   The robot moves toward the user with the beverage, but the motion planner fails and it cannot complete the serving attempt. The team says the drink was not located correctly relative to the table.

4. 6:27

   Locating the table

   The team explains that the robot uses a map and the same algorithm to account for table placement and rotation. It detects four points or corners to locate the table, whose position is otherwise predefined.

5. 8:01

   Hand limits and human-like motion

   In the discussion, the team says its analysis of the provided hand found limitations with some objects. They describe exploring tactile sensors and coupling degrees of freedom into synergies to make manipulation faster and more human-like.