Grasping and manipulating contact-sensitive objects (CSOs) \u2014 fruits, vegetables, fish fillets, plant cuttings, and fragile geological samples \u2014 is a common task across the agriculture and food industries, yet these objects are easily damaged. Too much force crushes them; too little force, or an inappropriate grasp configuration, drops them. Handling them safely requires two conditions to be met at once: a suitable grasp configuration, because a poor configuration may demand more force than the object can tolerate, leaving no safe grasp; and a suitable amount of force given that configuration. Robotic manipulation that satisfies both conditions \u2014 an appropriate grasp configuration and a suitable force \u2014 is defined as gentle robotic manipulation.
As agricultural automation and manufacturing sophistication rise, demand for gentle manipulation of CSOs is growing. In Australia, food waste imposes an estimated $36.6 billion annual burden on the economy, with a significant share driven by mechanical damage during manufacturing, processing, and distribution; this waste also generates greenhouse-gas emissions once sent to landfill. Soft grippers, compliant manipulators, and gentler grasping strategies have been proposed to reduce this loss, but progress is constrained by a more fundamental gap: tactile interaction data \u2014 the key signal for learning gentle manipulation \u2014 is scarce, expensive to collect, and weakly grounded in what the object itself experiences during contact. Learning systems are trained on indirect proxies \u2014 external vision, gripper-side tactile sensors, or simulation \u2014 that only approximate the true contact forces acting on the object. Without this grounding, learned behaviour generalises poorly and damage thresholds remain guesswork.
This project develops physically grounded, contact-centric learning for gentle robotic manipulation: rather than learning from raw sensor signals alone, robots learn representations of the physical contact dynamics \u2014 slip onset, compression, contact stability \u2014 that actually determine whether manipulation is gentle. Central to the approach are instrumented objects: sensor-embedded mock objects that measure contact from inside the grasped object (a first-object perspective) and replicate the physical properties of real CSOs. Acting as physical teachers, they supply ground-truth contact measurements that anchor learning in real physics, complement large-scale simulation, and provide a shared basis for evaluation. The project lays the foundations for gentle manipulation skills that transfer across objects, grippers, and sensors.<\/p>\n\n\n\n
Aim <\/p>\n\n\n\n
This project aims to develop physically grounded, contact-centric learning for gentle robotic manipulation, using instrumented objects and simulation to supply the ground-truth contact data that learning requires.<\/p>\n\n\n\n
Objectives <\/p>\n\n\n\n
The project spans two complementary pillars; the precise focus will be scoped with the supervisor to suit the candidate’s background and interests.<\/p>\n\n\n\n
Significance <\/p>\n\n\n\n
The proposed research will:<\/p>\n\n\n\n
Ideal Candidate <\/p>\n\n\n\n
We are seeking a self-motivated PhD candidate who:<\/p>\n\n\n\n
This project is open to International and Domestic applicants. <\/p>\n\n\n\n
Internship <\/p>\n\n\n\n
Through this project you will also have an internship opportunity.\u00a0There are internship opportunities with domestic robotics companies such as LYRO Robotics (https:\/\/lyro.com), which specialises in robotic systems for handling agricultural products, and Sydekick Robotics (https:\/\/www.sydekick.bot), a Queensland-based company building an AI-powered platform for creating dexterous, human-like robot manipulation skills.<\/p>\n\n\n\n
Scholarship <\/p>\n\n\n\n
If you are identified as the preferred candidate for this project, you may be considered for an RTP scholarship<\/a>. <\/p>\n\n\n\n Enquires and How to Apply <\/p>\n\n\n\n For enquires about this opportunity contact Dr Fangyi Zhang at\u00a0Fangyi.Zhang@curtin.edu.au<\/a><\/p>\n\n\n\n