• dual arm manipulation
• force controller
• human-like motion planning
• hybrid controller
• impedance controller
• sensorless force estimation
• vision

In scenarios involving complex object manipulation, certain tasks, such as handling
cumbersome objects, necessitate the collaboration of multiple robotic arms as a single
arm proves insufficient. In the context of a multi-manual system, it becomes imperative to establish a well-defined control policy aligning with the task objectives. This
thesis implements a modular control law tailored for uni-arm manipulation within the
framework of a multi-manual setup, focusing on tasks involving unknown objects.
Drawing inspiration from force and impedance control principles, a novel adaptation
policy is implemented to address inherent limitations observed in conventional methods, including coupled impedance and hybrid position/force control approaches. The
implementation ensures stable grasping before executing the manipulation task, incorporating human-like motion planning for a seamless and safe operation.in terms of object analysis, the main sensor setup is one RGB-D camera. The generated point cloud
is then rigorously filtered and evaluated to account for incomplete surface coverage.
This allows significant geometric parameters to be extracted, including the dimensions
of the bounding box and the relative contact point identification. In essence, the setup
for the experiment involves deploying two Franka manipulators in random configurations to simulate real-world variety. The formal workflow can be organized into three
main phases:
Pre-Contact Phase: This initial phase corresponds to a thorough evaluation of the
object with the aim to identify important contact areas that are required for further
manipulation activities,
Contact Phase: By employing impedance control strategies, the system coordinates appropriate motions in accord to the human like planner to approach and come
into contact with key locations on the object’s surface that have been previously spotted.
Post-Contact Phase: Transitioning into the hybrid control policy, and beginning
the object manipulation