Hybrid Control for Robotic Nut Tightening Task

TL;DR

This work addresses autonomous nut tightening with a serial manipulator by introducing a hierarchical planner that builds task-space keyframes followed by trajectory optimization, and a hybrid control scheme that switches between stiffness (position) control and force control in task space during contact-rich phases. The main contributions are a motion-primitive-based planning framework, a robust hybrid controller with a secondary objective, and an open-source implementation; simulations show the system is 14.5% faster and applies two orders of magnitude less contact force than a stiffness baseline. The approach demonstrates strong robustness to initial condition variance and maintains low computational cost, highlighting its potential for broader contact-rich manipulation tasks. The work paves the way for faster, safer fastening operations in industrial and home settings and provides a reusable software release for the robotics community.

Abstract

An autonomous robotic nut tightening system for a serial manipulator equipped with a parallel gripper is proposed. The system features a hierarchical motion-primitive-based planner and a control-switching scheme that alternates between force and position control. Extensive simulations demonstrate the system's robustness to variance in initial conditions. Additionally, the proposed controller tightens threaded screws 14% faster than the baseline while applying 40 times less contact force on manipulands. For the benefit of the research community, the system's implementation is open-sourced.

Figures (8)

• Figure 1: A schematic depiction of the considered task. An autonomous manipulation system that uses a parallel gripper and signals from F/T sensor to tighten a nut onto a bolt is proposed.
• Figure 2: The proposed autonomous manipulation system architecture for nut tightening task. The planner iterates through the approach, screw, retract stages to guide the robot's interaction with the manipuland. Hybrid control is applied during the screw stage and stiffness control is applied otherwise.
• Figure 3: The planning framework models explicitly the three stages of nut tightening task: a) approach, b) screw, c) retract. Note: the forward direction of the gripper (co-axial with the major axes of the gripper's fingers) coincides with the Y-axis of the gripper's model (green). The opacity of the keyframes encodes the precedence information: lower-opacity keyframes must be reached before higher-opacity ones. 30$^{\circ}$ turn plan is shown in b).
• Figure 4: The grasp selection problem is solved during the keyframe planning for the screw stage. a) A hexagonal manipuland allows for six distinct stable grasps. b) The grasp that is the most advantageous kinematically is selected.
• Figure 5: A visualisation of simulated manipulation sequence by the proposed system. The robot completes three nut tightening sequences, with grasp adjustments between the screwing interactions. Each sequence includes the approach, screw, retract stages. Green arrows indicate the directions and relative magnitudes of the contact forces. The initial and terminal states of the system are omitted for brevity.