Tactile-based Robotic Manipulation of Delicate Compliant Objects

Why it matters

Delicate object manipulation is common in food processing, manufacturing and logistics. Delicate objects are normally vulnerable to mechanical damage. Globally, about one billion tonnes (valued US$940 billion) of food produced for human consumption is wasted each year. In Australia, there is an estimated $20 billion lost to the Australian economy each year due to food waste. Around a quarter of the wastage occurs in manufacturing processing, where mechanical damage to food or packaging due to handling is a major factor, particularly for fruits, vegetables and root and tuber crops. Once disposed of in landfill, food waste contributes to greenhouse gas emissions. This project aims for techniques and solutions for more gentle manipulation of delicate objects to help reduce the mechanical damage, food waste and greenhouse gas emissions, and to help achieve higher levels of autonomy in the food processing, agriculture, manufacturing and logistics industries.

Overview

Grasping and manipulating delicate compliant objects such as fruits and fishes are common tasks in the agriculture and food industries. With the increasing demands for agricultural automation and the increasing sophistication of manufacturing, the need for more reliable compliant object grasping and manipulation is growing. However, the performance of existing robotic or automation solutions for compliant objects is inadequate – either too little or too much force is applied the object may be damaged by dropping or crushing. Although this damage can be reduced using specifically designed soft grippers or compliant mechanisms, applying the right amount of force on the object will help further reduce or avoid the damage. However, little attention has been paid to the gentleness (minimum and non-damaging force) of robotic manipulation. To fill the gap, this project will study systematic ways of performance assessment that take gentleness into account, seek solutions that balance speed, safety and object damage by leveraging off-the-shelf tactile sensors, and explore cutting-edge machine learning techniques for compliant robotic manipulation applications.