Clustering Federated Learning (CFL) promotes knowledge sharing by adaptively grouping similar clients while solving the Non-IID problem to provide personalized models with high generalization. However, there are several problems needed to be considered before the practical deployment of CFL: 1) Practical clients often have limited communication capabilities. 2) Dynamically evolving data distributions of clients may leads to unreasonable clustering. 3) Asynchronization will cause clients belonging to the same cluster to fall apart. The above three problems will reduce the convergence speed and damage the accuracy of the model. In this work, we propose a Dynamic Representation-driven Clustering Federated Learning framework (DReCFL) to solve the above three problems. Specifically, DReCFL replaces complex model parameters with data representation for communicating to reduce the communication pressure. In order to gain a reasonable clustering, DReCFL utilizes a Dynamic Client Fuzzy Clustering (DCFC) algorithm, we proposed, to adapt to evolving data distributions. Finally, DReCFL leverages an Adaptive Clustering Threshold (ACT) mechanism, we designed, to craft clustering thresholds based on training progress, ensuring the reliability of client clustering results under asynchrony. Extensive experimental results demonstrate that DReCFL efficiently adapts to the addition of new clients and changes in data distribution, while reducing communication overhead by 22.70%-83.33% compared to state-of-the-art (SOTA) personalized federated learning methods.

Clustered Federated Learning (CFL) effectively mitigates the negative impact of data heterogeneity among clients on model accuracy by grouping clients with similar data distributions for training. However, existing methods require communication of model parameters between clients and the server, leading to high communication costs. Additionally, current clustering strategies restrict the flexibility of cluster partitioning. To address these issues, we propose a Dynamic Representation Clustering for Federated Learning (DRCFL) algorithm. It decouples the model into a Representation Generation Module (RGM) and a Shared header (S-header). The former calculates client data representations, while the server clusters these representations and trains a shared header for each cluster. Clients only need to transmit data representations to the server, remarkably reducing communication costs. Additionally, the server re-clusters based on the most recent data representations in each training round, enhancing the flexibility of cluster partitions. In terms of communication cost, Extensive experimental results indicate that DRCFL achieves a reduction of 22.70%−83.33% compared to state-of-the-art (SOTA) personalized FL methods.

Automatic repair of mechanical structures would enable a robot to recover or improve functions after physical damage. Little work exists on real-world execution of automatic repair in robotic systems. State-of-the-art takes a modular approach where the robotic system is modular and a replacement module is available. However, the modular approach suffers from low granularity in repair even with tens of motors. In addition, there is a lack of quantitative evaluation of the effect of automatic repair on robot functionality. Here we propose a cooperative method for automatic repair in a robotic system. Our method is regeneration-based rather than module-based and does not assume availability of a replacement part. It integrates a fabrication process on the fly for robot structure regeneration. With a system that consists of a regenerating robot, a legged robot and a pre-engineered ribbon, we demonstrate end-to-end execution of automated repair of the legged robot's leg by the regenerating robot in 335 seconds. Experiments on repeatability show a 100% success rate for sub-processes such as positioning, leg fabrication, and legged robot disengagement and a 90% success rate for leg detachment. We quantify the effect of leg regeneration on mobility recovery and found a 90% recovery of forward speed, a 19.7% increase of peak power and a 9.3% reduction of cost of transport with a regenerated leg.

OpenRoACH is a 15-cm 200-gram self-contained hexapedal robot with an onboard single-board computer. To our knowledge, it is the smallest legged robot with the capability of running the Robot Operating System (ROS) onboard. The robot is fully open sourced, uses accessible materials and off-the-shelf electronic components, can be fabricated with benchtop fast-prototyping machines such as a laser cutter and a 3D printer, and can be assembled by one person within two hours. Its sensory capacity has been tested with gyroscopes, accelerometers, Beacon sensors, color vision sensors, linescan sensors and cameras. It is low-cost within $150 including structure materials, motors, electronics, and a battery. The capabilities of OpenRoACH are demonstrated with multi-surface walking and running, 24-hour continuous walking burn-ins, carrying 200-gram dynamic payloads and 800-gram static payloads, and ROS control of steering based on camera feedback. Information and files related to mechanical design, fabrication, assembly, electronics, and control algorithms are all publicly available on https://wiki.eecs.berkeley.edu/biomimetics/Main/OpenRoACH.

Automatic folding has drawn increasing attention in robotics research in the past ten years. The focus has been on folding two-dimensional (2D) sheets into three-dimensional (3D) structures, but little work has been done on how structures may be formed by folding ribbons. Here we propose the concept of robotic ribbon folding including a general workflow from shape design to ribbon folding and shape retention. We also propose a method to realize robotic ribbon folding on the macroscopic scale. The method consists of minimally engineered ribbons with patterned flexures, a folding robot, and a folding scheme that relates the orientation of flexures, the type of folds and the type of structural elements. By using this method we demonstrate robotic ribbon folding into 2D static structures such as triangles and squares, 3D static structures, and planar kinematic linkages such as a simple non-crossing four-bar mechanism. Burn-in result shows a four-bar mechanism with all bars' length of 5 cm could move for over 660 cycles.

Ribbon folding is a new approach to structure formation that forms higher dimensional structures using a lower dimensional primitive, namely a ribbon. In this paper, we present a novel algorithm to address path planning for ribbon folding of multi-link planar structures. We first represent the desired structure with a graph-based representation of edges and nodes. We then use graph theory to claim that for any object which is represented by a connected graph, there exists a continuous path which visits all of its edges. Finally, we develop a path planning algorithm that takes into account the physical constraints of the folding machine. The input is the desired planar structure, and the output is the optimal sequence of ribbon folds for creating that structure using the minimum number of folds. The results of this algorithm are successfully used to fold various planar structures.

A dragline-forming technology has been previously proposed to enable locomotion through an open-space where no solid surfaces present. The technology is intended for situations where payload requirements are unanticipated. In those situations, variability in dragline's diameter can minimize the use of material hence increase self-sufficiency of the robot. In a previous study, a robot was designed, prototyped and proven to be able to descend through an open-space by forming a thermoplastic dragline with a diameter range of 1.1-4.5 mm. However, the speed of locomotion was rather low due to the lack of an adequate control method for thermoplastic dragline formation. In this paper, models of mass flow and thermodynamics along dragline formation pathway are presented. The models are validated in a newly prototyped robot which forms a dragline continuously. Experiment results show that, when compared to the previous prototype and control method which consists of repeated sequences of discrete events, the speed of descending locomotion is significantly increased and reaches 12.0 cm/min.

The paper presents a new concept of locomotion for wheeled or legged robots through an object-free space. The concept is inspired by the behaviour of spiders forming silk threads to move in 3D space. The approach provides the possibility of variation in thread diameter by deforming source material, therefore it is useful for a wider coverage of payload by mobile robots. As a case study, we propose a technology for descending locomotion through a free space with inverted formation of threads in variable diameters. Inverted thread formation is enabled with source material thermoplastic adhesive (TPA) through thermally-induced phase transition. To demonstrate the feasibility of the technology, we have designed and prototyped a 300-gram wheeled robot that can supply and deform TPA into a thread and descend with the thread from an existing hanging structure. Experiment results suggest repeatable inverted thread formation with a diameter range of 1.1-4.5 mm, and a locomotion speed of 0.73 cm per minute with a power consumption of 2.5 W.

Despite many approaches proposed in the past, robotic climbing in a complex vertical environment is still a big challenge. We present here an alternative climbing technology that is based on thermoplastic adhesive (TPA) bonds. The approach has a great advantage because of its large payload capacity and viability to a wide range of flat surfaces and complex vertical terrains. The large payload capacity comes from a physical process of thermal bonding, while the wide applicability benefits from rheological properties of TPAs at higher temperatures and intermolecular forces between TPAs and adherends when being cooled down. A particular type of TPA has been used in combination with two robotic platforms, featuring different foot designs, including heating/cooling methods and construction of footpads. Various experiments have been conducted to quantitatively assess different aspects of the approach. Results show that an exceptionally high ratio of 500% between dynamic payloads and body mass can be achieved for stable and repeatable vertical climbing on flat surfaces at a low speed. Assessments on four types of typical complex vertical terrains with a measure, i.e., terrain shape index ranging from -0.114 to 0.167, return a universal success rate of 80%-100%.

The capability of extending body structures is one of the most significant challenges in the robotics research and it has been partially explored in self-reconfigurable robotics. By using such a capability, a robot is able to adaptively change its structure from, for example, a wheel like body shape to a legged one to deal with complexity in the environment. Despite their expectations, the existing mechanisms for extending body structures are still highly complex and the flexibility in self-reconfiguration is still very limited. In order to account for the problems, this paper investigates a novel approach to robotic body extension by employing an unconventional material called Hot Melt Adhesives (HMAs). Because of its thermo-plastic and thermo-adhesive characteristics, this material can be used for additive fabrication based on a simple robotic manipulator while the established structures can be integrated into the robot's own body to accomplish a task which could not have been achieved otherwise. This paper first investigates the HMA material properties and its handling techniques, then evaluates performances of the proposed robotic body extension approach through a case study of a “water scooping” task.

Robust climbing in unstructured environments is a long-standing challenge in robotics research. Recently there has been an increasing interest in using adhesive materials for that purpose. For example, a climbing robot using hot melt adhesives (HMAs) has demonstrated advantages in high attachment strength, reasonable operation costs, and applicability to different surfaces. Despite the advantages, there still remain several problems related to the attachment and detachment operations, which prevent this approach from being used in a broader range of applications. Among others, one of the main problems lies in the fact that the adhesive characteristics of this material were not fully understood fin the context of robotic climbing locomotion. As a result, the previous robot often could not achieve expected locomotion performances and “contaminated” the environment with HMAs left behind. In order to improve the locomotion performances, this paper focuses on attachment and detachment operations in robot climbing with HMAs. By systematically analyzing the adhesive property and bonding strength of HMAs to different materials, we propose a novel detachment mechanism that substantially improves climbing performances without HMA traces.

Vertical climbing on a variety of flat surfaces with a single robot has been previously demonstrated using vacuum suction, electrostatic adhesion, and biologically inspired approaches, etc. These methods generally have a low attachment strength, and it is not clear whether they can provide satisfactory attachment on vertical terrains with richer 3D features. Recent development of a climbing technology based on hot melt adhesives (HMAs) has shown its advantage with a high attachment strength through thermal bonding and viability to any solid surfaces. However, its feasibility for vertical climbing has only been proven on flat surfaces and with external energy supplies. This paper provides quantitative measurements for vertical climbing performance on five types of surfaces and terrains with a self-contained robot exploiting HMAs. We show that robust vertical climbing on multiple terrains can be achieved with reliable high-strength attachment.