Juan Rojas

Towards snap sensing

Automating snap assemblies is highly desirable but challenging due to their varied geometrical configurations and elastic components. A key aspect to automating snap assemblies is robot state estimation and corrective motion generation, here defined as snap sensing. While progress is being made, there are yet no robust systems that allow for snap sensing. To this end we have integrated a framework that consists of a control strategy and control framework that generalises to cantilever snaps of varying geometrical complexity. We have also integrated a robot state verification method (RCBHT) that encodes FT data to yield high-level intuitive behaviours and perform output verification. Optimisation procedures and Bayesian filtering have been included in the RCBHT to increase robustness and granularity. The system provides belief states for higher level behaviours allowing probabilistic state estimation and outcome verification. In this work, preliminary assembly failure characterisation has been conducted and provides insights into assembly failure dynamics. The results, though still in simulation, are promising as the framework has effectively executed cantilever snap assemblies and robust robot state estimation with parts of varying complexity in two different robotic systems.

Early failure characterization of cantilever snap assemblies using the PA-RCBHT

Failure detection and correction is essential in robust systems. In robotics, failure detection has focused on traditional parts assembly, tool breakage, and threaded fastener assembly. However, not much work has focused on sub-mode failure classification. This is an important step in order to provide accurate failure recovery. Our work implemented a novel failure characterization scheme for cantilever snap assemblies. The approach identified exemplars that characterized salient features for specific deviations from a nominal trajectory. Then, a rule based approach with statistical measures was used to identify failure and classify failure sub-modes. Failure sub-mode classification was evaluated by using a reliability measure. Our work classified failure deviations with 88% accuracy. Varying success was experienced in correlating failure deviation modes. Cases with only 1-deviation had 86% accuracy, cases with 2-deviations had 67% accuracy, and cases with 3 deviations had 55% accuracy. Our work is an important step in failure characterization of complex geometrical parts and serves as a stepping stone to enact failure recovery.

A swarm framework for teaching elementary addition operations

The advent of low-cost, functional robots has promoted swarm robotics research. However, an area of research that is yet untouched is how these types of robots could aid in teaching STEM subjects. In particular, there is little work concerning mathematics at the elementary level. The Kilobot robots size, cost, and functionality offers a good test-bed to explore its use as a pedagogical tool. This work presents a swarm framework to teach elementary mathematical addition operations to children. A state-based synchronous algorithm was used to enable robots to represent both operands and result digits in mathematical operations, all the while exploiting the robots sounds, motion, and illumination to enhance childrens learning factors. The system is demonstrated in the V-REP simulation environment to demonstrate the feasibility of the approach.

Cantilever snap assemblies failure detection using SVMs and the RCBHT

Failure detection plays an increasingly important role in industrial processes and robots that serve in unstructured environments. This work studies failure detection on cantilever snap assemblies, which are critical to industrial use and growing in importance for personal use. Our aim is to study whether an SVM can use a small set of features abstracted as behavior representations from the assembly force signature to accurately detect failure at different stages of the task. In this work, a linear SVM was embedded with abstracted behavioral features to classify failure detection in cantilever snap assembly problems. The approach was useful in detecting failure offline during early and late stages of the task. For early stages, low-abstraction behaviors sets performed better due to their granularity and local temporal nature. For late stage analysis, high-abstraction behaviors performed better due to their coarse and global representations.

Contextualized early failure characterization of cantilever snap assemblies

Failure detection and correction is essential in robust systems. In robotics, failure detection has focused on traditional parts assembly, tool breakage, and threaded fastener assembly. However, not much work has focused on classifying failure into various sub-modes. This is an important step in order to provide accurate failure recovery. Our work implemented a contextualized failure characterization scheme for cantilever snap assemblies. A rule based approach was used through which assemblies whose trajectories deviated from the normal approach trajectory were identified in the beginning of the task. We not only identified failure but also the failure type that occurred. The method identified exemplars that characterized salient features for specific deviations from the initial approach trajectory in the assembly task. A contact-state map was generated through sampling the contact space during training. Contextualized statistical measures were used to classify trials during the testing phase. Our work classified failure deviations with 88% accuracy. According to the statistic measures used, varying success was experienced in correlating failure deviation modes. Each case was analyzed using gaussian statistics and one and two standard deviations. Cases with trajectory deviations in one direction had {75%, 92%} accuracy, cases with deviations in two directions had {61%, 94%} accuracy, and cases with deviations in three directions had {69%, 100%} accuracy. Our work provides further insights into the early failure characterization of complex geometrical parts which will serve to implement failure recovery techniques in the face of significant and unexpected errors.

Gradient calibration for the RCBHT cantilever snap verification system

In this work a gradient calibration method was presented as part of the Relative-Change-Based-Hierarchical Taxonomy (RCBHT) cantilever-snap verification system and the Pivot Approach control strategy for the automation of cantilever-snaps. As part of a relative-change based force signal interpretation scheme, an effective gradient calibration process is needed to increase the RCBHTs system robustness. Prior to this work, all gradient classification schemes were derived on an intuitive trial and error basis. Statistical measures were used to derive contact and constant gradient thresholds in contextually sensitive ways. The method requires training assemblies to identify a minimum contact gradient which serves as a marker for all other gradient thresholds. Experimental procedures verified that our calibration method was effective. Assemblies with supervised successful outcomes were used in experimentation. The RCBHT assessed out assemblies as successful using the calibration method. Even two snaps that where classified falsely as unsuccessful when using a previously non-calibrated version of the RCBHT.

Assessment of a proprietary online smart-family-matching tool to reunite lost families

The presence of wars, civil strife, and natural disasters inevitably separate many families from their loved ones. Existing tools to help divided families find loved ones consist of basic meta-data searches that are often unable to return a match in the face of incomplete or inaccurate information. This paper presents a first-order analysis of a proprietary online system that used correlated meta-data, genealogical information, and single images to search and match lost relatives more effectively. Experimental results showed that the system reliably guaranteed the right match in the presence of complete and accurate information. The results also showed that when incomplete meta-data information is present along with genealogical information, false-positive matches occur within ones own family tree. Furthermore, with increased genealogical data, the likelihood of false-positives can decrease by almost 44%. Image processing tools did not improve the results of the match when only one image is used. A larger database of family pictures could improve the likelihood of finding the right match. This paper assessed the viability of the system as a tool for humanitarian organization that help refugees located loved ones. Based on the results, we recommend the integration of genealogical data into existing search systems used by humanitarian organizations.

Proposal of a shape adaptive gripper for robotic assembly tasks

Graphical Abstract This paper proposes a novel robotic gripper used for assembly tasks that can adaptively grasp objects with different shapes. The proposed hand has a combined structure between two kinds of shape adaptive mechanisms where one is the granular jamming and the other is a multi-finger mechanism driven by a single wire. Due to the effect of the two shape adaptive mechanisms, the pose of a grasped object does not change during an assembly operation. The proposed hand has four fingers where two are the active ones and the other two are the passive ones. The pose of the grasped object can be uniquely determined since the passive fingers are used to orient an object placed on a table before the active fingers are closed to grasp it. Assembly experiments of some kinds of parts are shown to validate the effectiveness of our proposed gripper.

A constraint-based motion control strategy for cantilever snap assemblies

Industrial snap assembly processes remain largely a manual task. Much of the research in snap assembly has not sought to design strategies and controllers according to the class of snap-fastener used. Three fastener types are used in manufacturing: cantilever, torsional, and annular snaps. As a first step in solving the snap automation problem, sought to devise a force control strategy that could effectively perform cantilever-snap assemblies for various degrees of complexity and one that could generate reproducible sensory-motor signals across trials as a basis to facilitate the future discrimination of force signals and enable a robot to reason about the assemblys task state. Our contribution is two-fold, a control strategy that: (i) exploits constraint-motion designs built into cantilever snap parts to more effectively complete the task, and (ii) a strategy that can be applied to cantilever-snap parts of growing varying complexity such as those containing, one, two, or four snaps. The control basis approach was used as a framework to design force controllers for the Pivot Approach. The frameworks modularity and scalability enables the flexible adaptation of force controllers to snaps of varying complexities and geometries. The Pivot Approach simulation results showed that the control strategy took advantage of hardware designs increasing the likelihood of successful insertions and yielding consistent sensory-motor signal patters across trials for the snap assembly. These results serve as foundational work to devise new signal interpretation methods to enable robot to reason about the assembly state and produce more fault tolerant behaviors.

A steering wheel manipulation scheme by an anthropomorphic humanoid robot in a constrained vehicle environment

In recent years, wheel turning tasks have become of interest in disaster response scenarios. Particularly valve-turning and wheel steering for tasks in which a humanoid drives a vehicle. Most research has focused in valve-turning with dual arm motions to turn the valve as much and as quickly as possible. Valve-turning compared to steer-wheel turning consist of very different constraints due to the tasks environment. Driving suffers from strong constraints in mobility, reachability, and visibility, all the while turning consists of accurate steering in sequences of various degrees, directions, and speeds. In this work we study how effectively the atlas humanoid robot can steer the driving wheel to commanded angles within the hard constraints of the task. A three step approach is suggested as part of the manipulation scheme: (i) A user guided task-level assignment for limited visibility to approximate the wheel plane; (ii) a discretization of the wheel to generate inverse kinematic (IK) solutions while constraining 4 DoFs; and (iii) a steering mechanism to turn the wheel in the presence of minimal feedback. Our results show that the approach effectively reaches commanded angles in the presence of hard constraints, with over 98.24% accuracy. This is a promising step as part of a greater approach to automate the process of vehicle driving.