At the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) on Nov 3rd, three papers on modular self-reconfigurable robots by Professor Tin Lun Lam were included in the IROS 2020. One of them was awarded the Best Paper Award on Robot Mechanisms and Design, a new category set up this year. CUHK-Shenzhen also became the first university in the Chinese mainland to win the IROS Award for Best Paper in the past decade.

As noted, the first authors of the three papers accepted by the conference are all Ph.D. students of CUHK-Shenzhen, all with Prof. Tin Lun Lam as the corresponding author. The team’s findings on modular self-reconfigurable robots were also featured by IEEE Spectrum, the flagship publication of IEEE.

About IROS

The IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) is one of the largest and most impacting robotics research conferences worldwide. Established in 1988 and held annually, IROS provides an international forum for the international robotics research community to explore the frontier of science and technology in intelligent robots and smart machines. Subject to COVID-19 restrictions, IROS 2020 has been moved from its original venue in Las Vegas to online and is free to the online public for the first time ever.

About the Project

Modular self-reconfigurable (MSR) robots are robots composed of a large number of repeated modules that can rearrange their connectedness to form a large variety of structures. Unlike traditional fixed-configuration robots that are designed for specific applications,  an MSR system can change its shape to suit the task or environment, and are highly adaptive and self-repairing. The ability to reconfigure allows a robot to disassemble and/or reassemble itself to form morphologies that are well-suited for a variety of given tasks.

However, MSR robots at home and abroad, even at the most advanced level, are still constrained by connectors, with the mutual positioning and motion planning functions of each module limited to structured scenarios; once they are put into unstructured environments, the rugged terrain, complex obstacles and lack of positioning information in the field will pose great challenges to the normal functions of the existing MSR systems.

Accordingly, this project investigates the key technologies for applying MSR robots to unstructured scenarios, and develops an MSR system that is free from laboratory scenarios and is highly adaptable to variable and uncertain environments, aiming to make breakthroughs in three aspects: 1) arbitrary connection point, 2) self-loaded mutual positioning techniques, and 3) collaborative path planning for adapting to obstacles. The three accepted papers in IROS 2020 were conducted in these three areas respectively.

About the Papers

1. FreeBOT: A Freeform Modular Self-reconfigurable Robot with Arbitrary Connection Point - Design and Implementation

First Author : Guanqi Liang, Year-2 Ph.D. student, SSE, CUHK-Shenzhen

Corresponding Author: Prof. Tin Lun Lam

IROS Best Paper Award on Robot Mechanisms and Design.

This paper proposes a novel modular self-reconfigurable robot (MSRR) "FreeBOT", which can be connected freely at any point on other robots. FreeBOT is mainly composed of two parts: a spherical ferromagnetic shell and an internal magnet. The connection between the modules is genderless and instant, since the internal magnet can freely attract other FreeBOT spherical ferromagnetic shells, and not need to be precisely aligned with the specified connector. This connection method has fewer physical constraints, so the FreeBOT system can be extended to more configurations to meet more functional requirements.

FreeBOT can accomplish multiple tasks although it only has two motors: module independent movement, connector management and system reconfiguration. FreeBOT can move independently on the plane, and even climb on ferromagnetic walls; a group of FreeBOT can traverse complex terrain. Many experiments have been conducted to test its function, which shows that the FreeBOT system has great potential to realize a freeform robotic system.

About the Author

Guanqi LIANG

Year-2 Ph.D. student, SSE, CUHK-Shenzhen

The first author of this paper is Guanqi Liang, a second year Ph.D. candidate at the School of Science and Engineering, The Chinese University of Hong Kong, Shenzhen. Prof. Tin Lun Lam is this research supervisor. Liang’s current research interests include modular self-reconfigurable robots and field robots.

Author's Note: This is the first work in my research journey and aims to design a novel modular self-reconfigurable robot (MSRR) that can effectively solve the problems of inefficient connection and limited configuration of existing MSRRs. I would like to thank Prof. Tin Lun Lam for introducing me to research and guiding me to explore the field of robotics. It was with his guidance that this paper could be included in this year’s IROS.

2. Robot-to-Robot Relative Pose Estimation based on Semidefinite Relaxation Optimization 

First Author: Ming Li, first-year Ph.D. student at SSE, CUHK-Shenzhen

Corresponding Author: Prof. Tin Lun Lam

In this paper, the 2D robot-to-robot relative pose (position and orientation) estimation problem based on egomotion and noisy distance measurements is considered. We address this problem using an optimization-based method, which does not require complicated numerical analysis while yields no inferior relative localization (RL) results compared to existing approaches. In particular, we start from a state-of-the-art method named square distances weighted least square (SDWLS), and reformulate it as a non-convex quadratically constrained quadratic programming (QCQP) problem. To handle its non-convex nature, a semidefinite programming (SDP) relaxation optimization-based method is proposed, and we prove that the relaxation is tight when measurements are free from noise or just corrupted by small noise. Further, to obtain the optimal solution of the relative pose estimation problem in the sense of maximum likelihood estimation (MLE), a theoretically optimal WLS method is developed to refine the estimate from the SDP optimization. Comprehensive simulations and well-designed experiments are presented for validating the tightness of the SDP relaxation, and the effectiveness of the proposed algorithm is highlighted by comparing it to the existing approaches.

Fig.1 The trajectories of robots on a two-dimensional plane

Fig. 2. A comparison of the three methods in terms of estimation accuracy based on real data

About the Author

Ming LI

First-year doctoral student, SSE, CUHK-Shenzhen

Ming Li, the first author of the paper, is a first-year Ph.D. student at The Chinese University of Hong Kong, Shenzhen, whose main research interests include robot positioning, multi-robot information fusion, and target optimization. His work is carried out supervised by Prof. Tin Lun Lam.

Author's Note: This work has helped me to further clarify the knowledge architecture of robot positioning, improve the knowledge required for positioning techniques, and improve my personal skills in theoretical analysis and scientific paper writing. In addition, special thanks to Prof. Tin Lun Lam for his valuable advice during the completion of the dissertation. His professional guidance was very helpful to the completeness and rigor of the dissertation.

3. An Obstacle-crossing Strategy Based on the Fast Self-reconfiguration for Modular Sphere Robots

First Author: Haobo Luo, First-year doctoral student, SSE, CUHK-Shenzhen

Corresponding Author: Prof. Tin Lun Lam

The fast obstacle-crossing strategy described in this paper consists of a number of self-reconfiguration processes. The path planning of each self-configuration process helps achieve an optimal value due to the parallel movement of many modules in one step, incorporated with the maximum gradient achieve fast reconfiguration. The final configuration of each self-configuration process is adjusted based on the chain connections of the spherical modules to adapt to the shape of the obstacles in the environment, thus allowing the robot to quickly climb over complex obstacles on rough ground.

The "fast speed" of the reconfiguration processes is due to a three-step minimization computation: first, the reconfiguration problem is modeled as a problem of minimizing the product of the Manhattan distance between each empty space in the final configuration and each module in the current configuration; second, each vacancy in the final configuration is assigned to the target module with the smallest Manhattan distance among the remaining movable modules; third, after two candidate actions are computed and the collision-causing candidate action is removed, the action that minimizes the Manhattan distance between the module and the target vacancy is selected.

The final configuration in each self-configuration needs to be adjusted from discrete to continuous space depending on the shape of the obstacle. The discrete 3D routes of the self-configuration algorithm can be smoothed out by calculating the increment for each step based on the rotation axis and rotation angle of the spherical module. In simulation experiments, the final configurations of each self-configuration process are able to adapt to a wide range of obstacles with the help of the spherical module's chain connection ability to adapt to complex surfaces.

About the Author

Haobo LUO

First Author: Haobo Luo, First-year doctoral student, SSE, CUHK-Shenzhen

Haobo Luo, the first author of this paper, is a first-year Ph.D. candidate at SSE, CUHK-Shenzhen specializing in computer information engineering. His research interests include intelligent optimization algorithms, graph theory, reinforcement learning algorithms, etc. He is currently responsible for the research on self-configuration algorithms and configuration optimization for modular self-configuration robots. His supervisor is Professor Tin Lun Lam.

Author's Note: This research field is very complicated and there are hundreds of schools of thought, each with their own merits. When I first started working in this field, I tried various approaches such as reinforcement learning and heuristic search to solve problems, but all of them were ineffective. This article is a preliminary exploration of this topic, which will help break through the next research challenges. I would like to thank Prof. Lam for his patience and guidance, and all of the team members for their cooperation.

About the Supervisor

Tin Lun LAM