|
Sensorless Force Control Algorithm based on Momentum Observer Technique |
|
Youngjun Joo, Hanbyeol Kim, Jonghun Park, Changeui Shin, Hoseong Kwak, Joonkeol Song, Chulho Shin(LG Electronics, Korea) |
| This paper presents a sensorless force control algorithm based on external force estimation technique for robotic manipulators. Instead of using F/T or joint torque sensors, we employ a momentum observer (MOB) for estimating external force, and it is used as feedback for force control. Stability and force control performance have been analyzed based on singular perturbation theory. In addition, a design guideline for applying MOB has been put together according to Lyapunov analysis. We validate performance of the proposed algorithm using experiments with industrial robotic manipulators. |
|
|
An Experimental Setup to Study the Performance of Flexure Mechanism |
|
Mohammmad Zubair, Bhivraj Suthar, Seul Jung(Chungnam National University, Korea) |
| Although the force analysis of flexure mechanism can be easily studied using the simulation environment, its realistic performance is very much required before it can be deployed to a real system. This paper presents an experimental setup to conduct the experiment on the flexure mechanism. A test-bed is built consisting of an actuator, a 6-axes load sensor, an accelerometer and a customized closure. A flexure mechanism is fabricated using 3D printer technology. A series of experiment was conducted to verify the simulation results and found it suitable for its deployment. |
|
|
Decomposed Q-learning for Non-prehensile Rearrangement Problem |
|
Hogun Kee, Dohyeong Kim, Songhwai Oh(Seoul National University, Korea) |
| We address a planar non-prehensile rearrangement task.
We model the problem as Multi-objective Markov decision processes (MOMDPs) and train the agent to learn object-specific Q-value functions according to the behavior of the robot arm by the individual objects.
The policy is based on the maximum strategy for using the learned object-wise Q-value functions.
Instead of the single Q-learning, the proposed object-wise decomposed Q-learning performed better. |
|
|
Trajectory Optimization with Geometry-Aware Singularity Avoidance for Robot Motion Planning |
|
Luka Petrovic, Filip Marić, Ivan Marković(University of Zagreb Faculty of Electrical Engineering and Computing, Croatia), Jonathan Kelly(University of Toronto Institute for Aerospace Studies, Canada), Ivan Petrović(University of Zagreb Faculty of Electrical Engineering and Computing, Croatia) |
| In this paper, we propose a cost function based on a novel geometry-aware singularity index and integrate it within a stochastic trajectory optimization framework for efficient motion planning with singularity avoidance.
We compare the proposed method with existing singularity-aware motion planning techniques, demonstrating improvement in common indices such as manipulability and dexterity and showcasing the ability of the proposed method to handle collision avoidance while retaining agility of the robot arm. |
|