POMDP-Guided Active Force-Based Search for Robotic Insertion

TL;DR

A novel search strategy that actively utilizes the information contained in the contact configuration and shows high efficiency is proposed that is based purely on the robot's proprioceptive sensing and does not need visual or tactile sensors.

Abstract

In robotic insertion tasks where the uncertainty exceeds the allowable tolerance, a good search strategy is essential for successful insertion and significantly influences efficiency. The commonly used blind search method is time-consuming and does not exploit the rich contact information. In this paper, we propose a novel search strategy that actively utilizes the information contained in the contact configuration and shows high efficiency. In particular, we formulate this problem as a Partially Observable Markov Decision Process (POMDP) with carefully designed primitives based on an in-depth analysis of the contact configuration's static stability. From the formulated POMDP, we can derive a novel search strategy. Thanks to its simplicity, this search strategy can be incorporated into a Finite-State-Machine (FSM) controller. The behaviors of the FSM controller are realized through a low-level Cartesian Impedance Controller. Our method is based purely on the robot's proprioceptive sensing and does not need visual or tactile sensors. To evaluate the effectiveness of our proposed strategy and control framework, we conduct extensive comparison experiments in simulation, where we compare our method with the baseline approach. The results demonstrate that our proposed method achieves a higher success rate with a shorter search time and search trajectory length compared to the baseline method. Additionally, we show that our method is robust to various initial displacement errors.

Figures (11)

• Figure 1: An overview of the proposed method. (a) Illustration of the search problem. The robot needs to align the peg with the hole. (b) Illustration of the proposed search strategy. Our search strategy actively applies force to exploit the information contained in the contact configuration and locate the hole's position. (c) Search paths of different search methods. From left to right: blind search with Spiral curve, blind search with Lissajous curve, and our method. The hole and peg centers are represented as black and red circles, respectively. Our search method can directly align the peg with the hole along a straight line.
• Figure 2: Three types of contact configuration between peg and hole. (a) No overlapping (b) Partially overlapping (c) Perfect alignment
• Figure 3: The planar problem and moment labeling. (a) The simplified planar problem. (b) Moment labeling of a contact wrench. (c) Moment labeling of a friction contact, where $\theta = 2\arctan(\mu)$ and $\mu$ is the friction coefficient. The "+" region consists of points where all contact wrenches only apply positive moment, the "-" region consists of points where all contact wrenches only apply negative moment, and the white region consists of points where the contact wrenches can apply both positive and negative moment.
• Figure 4: Moment labeling analysis of the planar problem. The red arrow represents the applied force on the peg. (a) Apply a downward force at the right end of the peg. The applied force cannot be compensated by the contact wrenches. The peg will tilt rightward. (b) Apply a downward force at the left end of the peg. The applied force can be compensated by the contact wrenches. The peg will keep static.
• Figure 5: Static stability analysis of the 3D contact configuration between the peg and the hole. The support polygon is the shadowed region. The intersection line is the blue line with two red-point ends. (a) If the projection of the downward applied force is inside the support polygon, the peg will keep static. (b) If the projection of the downward applied force is outside the support polygon, the peg will tilt around the intersection line.

Theorems & Definitions (1)

• Remark 1