R-3 Repository :: Browsing by Author "Kavraki, Lydia E"

Browsing by Author "Kavraki, Lydia E"

Now showing 1 - 19 of 19

A Constraint and Sampling-Based Approach to Integrated Task and Motion Planning
(2014-09-30) Prabhu, Sailesh Naveena; Chaudhuri, Swarat; Kavraki, Lydia E; McLurkin, James; Vardi, Moshe V
This thesis tackles the Integrated Task and Motion Planning (ITMP) Problem. The ITMP problem extends classical task planning with actions that require a motion plan. The agent seeks a sequence of actions and the necessary motions to achieve the goal. The user partially specifies the task plan by providing the actions' known parameters. An SMT solver, then, discovers values for the unkown parameters that satisfies constraints requiring the task plan to achieve the goal. The SMT solver utilizes an annotated Probabilistic Roadmap (PRM) to query for motion planning information. A sampling algorithm generates the PRM's vertices to permit a mobile manipulator to grasp numerous object configurations. Each iteration samples several base configurations and adds a base configuration to the PRM that increases the object configurations grasped from its vertices. Our results indicate that increasing the samples per iteration improves the probability the SMT solver discovers a satisfying assignment without adversely affecting the resulting task plan.
A Constraint-Based Approach to Reactive Task and Motion Planning
(2016-01-26) Wang, Yue; Chaudhuri, Swarat; Kavraki, Lydia E; Vardi, Moshe Y
This thesis presents a novel and scalable approach for Reactive Task and Motion Planning. We consider changing environments with uncontrollable agents, where the robot needs a policy to respond correctly in the infinite interaction with the environment. Our approach operates on task and motion domains that combine actions over discrete states with continuous, collision-free paths. We synthesize a policy by iteratively verifying and searching for a policy candidate. For efficient verification, we employ Satisfiability Modulo Theories (SMT) solvers using a new extension of proof rules for Temporal Property Verification. For efficient policy search, we apply domain-specific heuristics to generalize verification failures. Furthermore, the SMT solver enables quantitative specifications such as energy limits. We benchmark our policy synthesizer in a mobile manipulation domain, showing that our approach offers better scalability compared to a state-of-the-art robotic synthesis tool in the tested benchmarks and demonstrating order-of-magnitude speedup from our heuristics.
A SNP Calling And Genotyping Method For Single-cell Sequencing Data
(2015-04-23) Zafar, Hamim; Nakhleh, Luay K.; Kavraki, Lydia E; Jermaine, Chris M; Chen, Ken
In this thesis, we propose a single nucleotide polymorphism (SNP) calling and genotyping algorithm for single-cell sequencing data generated by the recently developed single-cell sequencing (SCS) technologies. SCS methods promise to address several key issues in cancer research which previously could not be resolved with data obtained from second generation or next-generation sequencing (NGS) technologies. SCS has the power to resolve cancer genome at a single-cell level and can characterize the genomic alterations that might differ from one cell to another. SNPs are the most commonly occurring genomic variations that alter the gene functions in cancer. Several methods exist for calling SNPs from NGS data. However, these methods are not suitable in the SCS scenario because they do not account for the various amplification errors associated with the SCS data. As a result, the existing SNP calling methods perform poorly, producing a large number of false positives when applied on SCS data. To the best of our knowledge, no SNP calling method exists that is specifically designed for SCS data. Our SNP calling algorithm is specifically designed for SCS data and the underlying statistical model deals with the inherent errors of SCS like allelic dropout, high bias for C : G > T : A and other amplification errors. This results in ~50% reduction in the number of false positives and ~30% increase in precision in calling SNPs as compared to GATK, a state-of-the-art SNP calling method for NGS data. Our algorithm also employs an improved genotyping method to properly genotype the individual cells by avoiding the sequencing errors (e.g., base calling error). Our method is the first SCS-specific SNP calling method and it can be used to characterize the SNPs present in individual cancer cells. Potentially, it can be applied as a first step in the genealogical analysis of tumor cells for tracing the evolutionary history of a tumor.
A Unifying Framework for Constrained Sampling-Based Planning
(2017-12-01) Kingston, Zak; Kavraki, Lydia E
Complex robots with many degrees-of-freedom (e.g., humanoids, mobile manipulators) have been increasingly applied to achieve tasks in fields such as disaster relief or spacecraft logistics. Finding motions for these systems autonomously is necessary if they are to be applied in unstructured environments not known a priori, as they must compute motions on-the-fly. Sampling-based motion planning algorithms have been shown to be effective for finding motions for high-dimensional systems such as these. However, the problems these robots face typically take the form of tasks with constraints, which limit the valid motions a robot can take (e.g., turning a valve about its axis, carrying a tray with both arms, etc.). Incorporating constraints while planning introduces significant challenges, as constraints induce a lower-dimensional manifold of constraint-satisfying configurations within the robot’s configuration space. The lower-dimensional structure of the manifold throws a wrench into the basic operation of a sampling-based planner, necessitating a constraint methodology to provide a means for the planner to satisfy constraints. Within the literature, many constrained sampling-based motion planning methods have been proposed for sampling-based planning with constraints. Each of these methods introduces a constraint methodology of their own to tackle the issues raised when considering constraints. This thesis organizes the menagerie of constraint methodologies along of a spectrum, cataloged by the amount of bookkeeping and computation used to approximate the manifold of constraint-satisfying configurations. Notably, previous constrained sampling-based methods augment a single sampling-based algorithm with their constraint methodology to create a bespoke planner. This thesis presents a general framework for sampling-based motion planning with geometric constraints, unifying prior works by approaching the constrained motion planning problem at a higher level of abstraction. The framework decouples the constraint methodology from the planner’s method for exploration by presenting the constraint-induced manifold as a configuration space to the planner, hiding details of the constraint methodology behind the space’s primitive operations. Three constraint methodologies from the literature are emulated within the framework. The framework is demonstrated with a range of planners using the three emulated constraint methodologies in a set of simulated problems. Results show the advantages decoupling brings to constrained sampling-based planning, with novel combinations of planners and constraint methodologies surpassing emulated prior works. The framework is also easily extended for novel planners and constraint spaces.
Algorithms for Scalable Structural Analysis of Class I Peptide-MHC Systems
(2020-04-22) Abella, Jayvee Ralph; Kavraki, Lydia E; Nakhleh, Luay K
Peptide-MHC (pMHC) complexes are central components of the immune system, and understanding the mechanism behind stable pMHC binding will aid the development of immunotherapies. Stable pMHC binding can be assessed through an analysis of structure, which contain information on the atomic interactions present between peptide and MHC. However, a large-scale analysis of pMHCs is difficult to perform, due to the lack of available structures as well as fact that pMHCs are large molecular systems with slow timescales. This thesis presents a set of approaches developed to deliver scalable structural analysis of Class I pMHC systems. First, we present APE-Gen, a fast method for generating ensembles of bound pMHC conformations. Next, we present a structure-based classifier using random forests for predicting stable pMHC binding. Finally, we present a simulation framework for generating a Markov state model of the full binding dynamics for a given pMHC system using a combination of umbrella and adaptive sampling. This work pushes the capability of computational methods for the structural analysis of pMHCs, leading to structural insight that can guide the understanding of pMHC binding.
Charge-based interactions through peptide position 4 drive diversity of antigen presentation by human leukocyte antigen class I molecules
(Oxford University Press, 2022) Jackson, Kyle R; Antunes, Dinler A; Talukder, Amjad H; Maleki, Ariana R; Amagai, Kano; Salmon, Avery; Katailiha, Arjun S; Chiu, Yulun; Fasoulis, Romanos; Rigo, Maurício Menegatti; Abella, Jayvee R; Melendez, Brenda D; Li, Fenge; Sun, Yimo; Sonnemann, Heather M; Belousov, Vladislav; Frenkel, Felix; Justesen, Sune; Makaju, Aman; Liu, Yang; Horn, David; Lopez-Ferrer, Daniel; Huhmer, Andreas F; Hwu, Patrick; Roszik, Jason; Hawke, David; Kavraki, Lydia E; Lizée, Gregory
Human leukocyte antigen class I (HLA-I) molecules bind and present peptides at the cell surface to facilitate the induction of appropriate CD8+ T cell-mediated immune responses to pathogen- and self-derived proteins. The HLA-I peptide-binding cleft contains dominant anchor sites in the B and F pockets that interact primarily with amino acids at peptide position 2 and the C-terminus, respectively. Nonpocket peptide–HLA interactions also contribute to peptide binding and stability, but these secondary interactions are thought to be unique to individual HLA allotypes or to specific peptide antigens. Here, we show that two positively charged residues located near the top of peptide-binding cleft facilitate interactions with negatively charged residues at position 4 of presented peptides, which occur at elevated frequencies across most HLA-I allotypes. Loss of these interactions was shown to impair HLA-I/peptide binding and complex stability, as demonstrated by both in vitro and in silico experiments. Furthermore, mutation of these Arginine-65 (R65) and/or Lysine-66 (K66) residues in HLA-A*02:01 and A*24:02 significantly reduced HLA-I cell surface expression while also reducing the diversity of the presented peptide repertoire by up to 5-fold. The impact of the R65 mutation demonstrates that nonpocket HLA-I/peptide interactions can constitute anchor motifs that exert an unexpectedly broad influence on HLA-I-mediated antigen presentation. These findings provide fundamental insights into peptide antigen binding that could broadly inform epitope discovery in the context of viral vaccine development and cancer immunotherapy.
Combining Discrete and Continuous Reasoning for Robot Motion Planning in Complex Domains
(2016-04-25) Luna, Ryan James; Kavraki, Lydia E
Robots are being employed not only for assembly tasks, but also in domains like healthcare, mining, and around the home. As robots become more capable, effective planning becomes critical since optimization-based techniques often fail to perform the kinds of deliberation required for complex systems. This thesis proposes new techniques for solving two instances of motion planning: path planning for high-dimensional manipulators and optimal task planning under uncertainty. Although each problem differs in its applications, the proposed solutions each build upon the notion of mixing discrete and continuous search to better inform the overall planning process. Sampling-based algorithms are known for their ability to quickly compute paths for robotic manipulators. The same algorithms are also notorious for poor quality solution paths, especially as the search space grows. This work demonstrates that, when planning for a robotic manipulator, high-quality paths can be computed in times competitive with the state-of-the-art by guiding a series of points on the robot through a decomposition of the workspace. The proposed algorithm scales to high-dimensional manipulators with dozens of joints by focusing the search to subsets of joints that affect motion in promising portions of the workspace. Simulated experiments on Robonaut2 and planar kinematic chains demonstrate the scalability and quality of the approach. Planning under uncertainty requires computing a policy over the robot's state space, a computationally intense task. Instead of solving a continuous Markov decision process (MDP) directly, this work proposes factoring the MDP} over discrete regions of the space for tractability. Locally optimal policies are constructed within each region to navigate the system to each adjacent region, and the set of regions and local policies are combined to form a bounded-parameter MDP where an optimal policy is computed by selecting one local policy for each region. Significant computational savings are realized by reasoning only over discrete regions, allowing the framework to compute policies that satisfy complex task specifications for an uncertain system. The bounded-parameter MDP can also be patched at runtime to compute policies in uncertain environments. Simulations demonstrate the proposed method computes an optimal policy in orders-of-magnitude faster time compared to existing techniques.
EnGens: a computational framework for generation and analysis of representative protein conformational ensembles
(Oxford University Press, 2023) Conev, Anja; Rigo, Mauricio Menegatti; Devaurs, Didier; Fonseca, André Faustino; Kalavadwala, Hussain; de Freitas, Martiela Vaz; Clementi, Cecilia; Zanatta, Geancarlo; Antunes, Dinler Amaral; Kavraki, Lydia E
Proteins are dynamic macromolecules that perform vital functions in cells. A protein structure determines its function, but this structure is not static, as proteins change their conformation to achieve various functions. Understanding the conformational landscapes of proteins is essential to understand their mechanism of action. Sets of carefully chosen conformations can summarize such complex landscapes and provide better insights into protein function than single conformations. We refer to these sets as representative conformational ensembles. Recent advances in computational methods have led to an increase in the number of available structural datasets spanning conformational landscapes. However, extracting representative conformational ensembles from such datasets is not an easy task and many methods have been developed to tackle it. Our new approach, EnGens (short for ensemble generation), collects these methods into a unified framework for generating and analyzing representative protein conformational ensembles. In this work, we: (1) provide an overview of existing methods and tools for representative protein structural ensemble generation and analysis; (2) unify existing approaches in an open-source Python package, and a portable Docker image, providing interactive visualizations within a Jupyter Notebook pipeline; (3) test our pipeline on a few canonical examples from the literature. Representative ensembles produced by EnGens can be used for many downstream tasks such as protein–ligand ensemble docking, Markov state modeling of protein dynamics and analysis of the effect of single-point mutations.
General Algorithms for the Time-Optimal Trajectory Generation Problem
(2017-01-25) Butler, Stephen D; Kavraki, Lydia E
As we ask robots to perform ever more sophisticated tasks, we must develop new algorithms capable of planning motions for these robots to follow. Current state-of-the-art humanoids and mobile manipulators generally operate under quasi-static assumptions. Ignoring dynamics during motion planning leads to highly sub-optimal trajectories. In turn, sub-optimal motion leads to higher requirements on robotic hardware such as increased weight, power consumption, and cost. Kinodynamic motion planning, which couples path planning with the dynamics of the robot, can create dynamically optimal trajectories. Unfortunately, kinodynamic motion planning is still intractable in the general case for high degree of freedom systems since including dynamics greatly increases the complexity of the motion planning search problem. This thesis presents two new algorithms which represent important steps toward closing the gap in general and efficient kinodynamic motion planning. The first algorithm solves the time-optimal path parameterization problem. The second algorithm can, given some range of starting velocities, generate all possible time-optimal trajectories for a given path by computing admissible velocity intervals (AVI). Solving the AVI problem generally and robustly is the first step toward a new generation of efficient kinodynamic motion planners which utilize the admissible velocity propagation (AVP) method. The algorithms presented improve on previous approaches by generically handling dynamics constraints. Through an exhaustive but efficient search, these new algorithms also eliminate the need for expert heuristics, which in prior methods were required to create the trajectory segments which form the final solution trajectory. The new algorithms presented in this thesis are simulated on a WAM arm with a Barret hand subject to dynamics constraints on joint torque, joint velocity, momentum, and end-effector velocity. Both algorithms are compared with a state-of-the-art alternative approach which they outperform in run-time and success rate.
Improving the interpretation of metabolic pathfinding results with clustering and compound hubs
(2016-04-27) Kim, Sarah Michelle; Kavraki, Lydia E
Knowledge on metabolic networks across species can be utilized to help address many challenges in biotechnology, including metabolic engineering. Large-scale annotated metabolic databases, such as KEGG and MetaCyc, provide a wealth of information to researchers designing novel biosynthetic pathways. However, many metabolic pathfinding tools that assist in identifying possible solution pathways fail to facilitate the interpretation of these pathway results. This work begins to address this problem by examining the performance of standard clustering algorithms on results produced by a popular metabolic pathfinding algorithm and suggesting the use of compound ”hubs” for examining the produced results. To address the first point, we assessed the ability of standard clustering method to expertly group pathways. Three standard clustering methods (hierarchical, k-means, and k-medoids) along with three pair-wise distance measures (Levenshtein, Jaccard, and n-gram) were used to group lysine, isoleucine, and 3-hydroxypropanoic acid (3-HP) biosynthesis pathways produced by a recent metabolic finding algorithm. The quality of the resulting clusters were quantitatively evaluated against expected pathway groupings taken from theliterature. Hierarchical clustering and Levenshtein distance appeared to best match external pathway labels across the three biosynthesis pathways but results suggest that grouping pathways with more complex underlying topologies may require more tailored clustering methods. In summary, the clustering of pathways proved much more nuanced than excepted due to the various intricacies of computed paths and several ways of getting between two compounds conserving the same number of atoms. To address the second point, we investigate the use of “hub” compounds. Hub compounds were selected by metabolic experts among compounds with a large number of in-degree reactions. An analysis of our results shows that hub compounds are common in the pathfinding results but that themselves alone cannot be used to cluster pathways. Our observations give rise to a new proposed method that will compute pathways between input and output compounds by using a precomputed a lookup table for pathways between the most well connected compound hubs in the metabolic network. The ultimate goal of precomputing the lookup table is to reduce search space while still obtaining most, if not all, pathway results found by the original search algorithm. We provide evidence that this is a promising direction for future research and can yield results that are more easily interpreted and refined by users.
Learned Heuristics for Task and Motion Planning with a Fixed Tabletop Manipulator
(2019-04-03) Wells, Andrew Marshall; Kavraki, Lydia E
As the physical capabilities of robots increase, so do the potential benefits of autonomous operation. Planning is a central component of an autonomous robotic system. Many practical robotics applications (e.g., setting a table) require reasoning about discrete actions (e.g., placing one plate at each seat) as well as continuous motions (e.g., moving a manipulator). These types of problems form the domain of Task-Motion Planning (TMP). They pose challenges for traditional task planners, as searching continuous, high-dimensional spaces typical of high-DOF manipulators is largely intractable for current discrete search techniques. They are challenging for typical motion planners because the discrete nature of the problems means solutions require sampling on subspaces of measure zero. Additionally, there is a natural hierarchy to the problem that we would like to exploit. As Garret et al. say: “We almost certainly do not want to decide whether to get the frying pan or the steak next by sampling configurations of the robot and the kitchen and testing for paths between them.” This combined task-motion planning problem is not a new research area. The 1971 paper by Fikes and Nilsson introduced a framework for solving such problems that is very similar to what one can find in a modern TMP solver. One notable point, or omission, is an assumption that calculating and representing motion feasibility will be an easy problem. For example they include a problem where a robot must jump onto a box to reach a light switch. They focus on how the robot knows this action is required without mention of how the robot can know if this action is feasible. Forty-seven years later, we still have no general way to determine whether such an action is feasible. Calculating motion feasibility and representing this feasibility to a task planner is still a core problem in TMP research. This thesis presents a new approach to this problem in limited domains, namely tabletop manipulation with a fixed robot arm. For such problems, we show that it is possible to learn a motion feasibility classifier and use it as a heuristic to guide the search for a task-motion plan. The learned heuristic guides the search towards feasible motions and thus reduces the total number of motion planning attempts. A critical property of our approach is the ability to provide robust planning in diverse scenes. We train the classifier on minimal exemplar scenes and then use principled approximations to apply the classifier to complex scenarios in a way that minimizes the effect of errors. By combining learning with planning, our heuristic yields order-of-magnitude run time improvements in diverse tabletop scenarios. Even when classification errors are present, properly biasing our heuristic ensures we will have little computational penalty.
Machine Learning-Based Prediction of Sites of Metabolism in Drugs: Exploring Feature Extraction Methods on Molecular Graphs
(2020-04-24) Mitchell, Nicole Elyse; Kavraki, Lydia E
Drug metabolism studies are a critical component of the drug design process. Metabolism of some drugs can lead to diminished therapeutic efficacy or even toxicity. The stability of a drug is expressed by the atoms, called Sites of Metabolism (SOMs), which undergo structural changes when that drug interacts with a metabolizing enzyme. Computationally predicting these metabolically labile atoms early on in the drug development process will enable pharmaceutical chemists to design molecules with favorable metabolic properties. A number of in silico methods have been developed for identifying SOMs, with a recent focus on machine learning due to its computational efficiency over structural modeling. Machine learning techniques classify atoms as SOMs based on feature vector representations. Existing approaches rely upon expert knowledge and often expensive experiments to engineer fixed atom descriptors with extensive sets of experimentally-derived attributes. However, models based upon learned instead of fixed representations have proven promising in other chemoinformatics tasks. Seeing molecules as attributed graphs, where atoms correspond to nodes and bonds correspond to edges, the SOM prediction problem can be formulated as a node classification task. We compare two methods of extracting node features from molecular graphs: a standard fingerprint generation strategy used by existing SOM prediction methods, which constructs task-agnostic node descriptors; and an unexplored approach based on a graph convolutional neural network, which learns task-specific node encodings. Both methods take into account the node attributes and graph connectivity to generate descriptive atom representations. We experiment with factors that can influence the performance of both feature extraction methods on a dataset commonly used in the literature for predicting SOMs. Despite the fact that the graph convolution approach requires more data and has more parameters to tune, we have achieved comparable performance between the two methods. Given enough data, we believe the graph convolution approach may reliably achieve improved performance over the fingerprint generation strategy. Our results indicate that the graph convolution approach can outperform the fixed fingerprint generation strategy when starting from molecular graphs that are not initialized with rich electro-chemical properties, demonstrating how learned representations could replace the need for expert-derived features for SOM prediction. Our results also illustrate the importance of tuning the feature extraction method to the metabolizing enzyme of interest.
Molecular docking and structural analysis for applications in biomedicine
(2023-04-12) Hall-Swan, Sarah; Kavraki, Lydia E
The discovery of new drugs and treatments can be facilitated by developing in silico tools. These new methods can guide in vitro experiments and elucidate immune mechanisms via sequence and structural analysis of biomolecules. For immunotherapy treatments, of particular interest are peptide-HLA class I (pHLA) complexes and T-cell receptors (TCRs), which play a crucial role in the immune response against viral infections and cancer. T-cell cross-reactivity, the ability of a single TCR to bind and respond to multiple pHLAs, is a significant aspect of T-cells. Predicting T-cell cross-reactivity can aid in the development of safer cancer immunotherapies and more effective viral vaccines. To this end, we first present PepSim, a novel scoring method that predicts T-cell cross-reactivity based on pHLA similarity. Our method, which is also implemented in a web-based tool, accurately distinguishes between cross-reactive and non-cross-reactive pHLAs across multiple datasets and can be utilized with any class I peptide-HLAs. Next, we leverage PepSim to identify potential vaccine targets against SARS-CoV-2 that may be cross-reactive with other SARS-CoV-2 peptides, thereby offering protection against numerous viral variants. Furthermore, we have created DINC-COVID, an ensemble docking tool that facilitates the development of SARS-CoV-2 drug therapies by taking into account ligand and receptor flexibility and generating plausible binding modes for SARS-CoV-2 proteins.
Retrieval-Based Learning for Efficient High-DoF Motion Planning
(2023-04-20) Chamzas, Constantinos; Kavraki, Lydia E; Shrivastava, Anshumali
Sampling-based planners are effective in many real-world applications such as robotics manipulation, navigation, and even protein modeling. However, in the absence of any prior information, sampling-based planners are forced to explore uniformly or heuristically, which can lead to degraded performance. One way to improve performance is to use prior knowledge of environments to adapt the sampling strategy. In this thesis, we propose Pyre, a retrieval-based framework for learning sampling distributions for high-dimensional motion planning. Pyre decomposes the workspace into local primitives, memorizing local experiences by these primitives in the form of local samplers, and storing them in a database. Pyre synthesizes a global sampler by learning to retrieve local experiences relevant to the given situation. This method transfers knowledge between diverse environments that share local primitives and significantly speeds up planning. Pyre learns incrementally and significantly improves performance with only a handful of examples, achieving better generalization over diverse tasks and environments as compared to prior approaches. We demonstrate the effectiveness of Pyre on 2D and 3D environments as well as with a simulated and real Fetch Robot tasked with challenging pick-and-place manipulation problems. Additionally, in this thesis, we also contribute MotionBenchMaker, an open-source benchmarking tool for realistic robot manipulation problems. It is an extensible, easy-to-use tool that allows users to both generate datasets and benchmark them by comparing motion planning algorithms.
Robot Manipulation Planning Under Linear Temporal Logic Specifications
(2015-12-07) He, Keliang; Kavraki, Lydia E; Vardi, Moshe Y; Chaudhuri, Swarat
Automated planning for manipulation tasks is highly desirable, for it enables robot manipulators to be used by none robotics experts. This thesis presents one approach to solving manipulation planning for tasks expressed in linear temporal logic (ltl). This approach is based on the synergistic framework, which provides probabilistic completeness guarantees. Even though the synergistic framework has shown to work well for planning for ltl tasks in the navigation domain, it lacked an abstraction that can capture the high dimensionality of manipulation. This thesis enables manipulation planning using the synergistic framework by introducing a manipulation abstraction and modifying the interaction between task and motion planning in the framework. The modified framework is shown to be effectively in case studies in both simulation and physical systems. The case studies also show that the synergistic framework plans for manipulation problems more effective using the manipulation abstraction in comparison with a naive abstraction.
Scaling Up Robotics-Inspired Conformational Sampling Algorithms
(2016-12-01) Abella, Jayvee Ralph; Kavraki, Lydia E
The ability to efficiently sample a protein’s conformational space allows one to understand how a protein may interact with different partners. Algorithms from sampling-based robot motion planning have been used for conformational sampling of small-sized systems. These algorithms keep track of “coverage” in conformational space based on what has been sampled and aim to intelligently perturb the protein’s degrees of freedom to bias search in less densely explored areas of conformational space. However, these algorithms were not designed for large proteins or complexes. These algorithms depend heavily on defining useful perturbation strategies, which is a very difficult task for large proteins because such systems are typically more constrained and exhibit complex motions. Additionally, conformational sampling generally becomes a harder problem as the size of the considered system increases, so these algorithms need to take advantage of significant computational resources when needed. This thesis describes SIMS 2.0, a new framework for conformational sampling built from prior work called the Structured Intuitive Move Selector (SIMS). We introduce an automated construction of perturbation strategies derived from B-factors, secondary structure, and rigidity analysis. We also introduce a new algorithm for conformational sampling that can take advantage of large-scale computational resources while still keeping the geometric reasoning that robotics-inspired algorithms excel at. This work pushes the limits of the size of systems that can be studied by robotics-inspired conformational sampling.
Swarm Robotics: Measurement and Sorting
(2015-04-24) Zhou, Yu; McLurkin, James D.; Kavraki, Lydia E; Chaudhuri, Swarat
To measure is an important ability for robots to sense the environment and nearby robots. Although camera, laser, and ultrasonic provide very accurate measurements, they are expensive and not scalable for large swarm of low-cost robots. The r-one robot designed at Rice University is equipped with infrared transmitters and receivers, which are designed for remote control and are very inexpensive in mass production. They are a good solution for short-range communication, since the signal attenuates at about 1 to 2 meters with appropriate voltage. This work describes my results in using them to measure bearing, orientation, and distance between nearby robots. However, infrared receivers are not designed for this kind of use, so I present a variable transmit power approach to allow useful and efficient local geometry measurements. With the ability to measure bearing and distance, I am able to solve the problem of sorting a group of n robots in a two-dimensional space. I want to organize robots into a sorted and equally-spaced path between the robots with lowest and highest label, while maintaining a connected communication network throughout the process. I begin with a straightforward geometry-based version of sorting algorithm, and point out there are many difficulties when communication range becomes limited. Then I describe a topology-based distributed algorithm for this task. I introduce operations to break the symmetry between minimum and maximum, in order to keep time, travel distance, and communication costs low without using central control. I run a set of algorithms (leader election, tree formation, path formation, path modification, and geometric straightening) in parallel. I show that my overall approach is safe, correct, and efficient. It is robust to population changes, network connectivity changes, and sensor errors. I validate my theoretical results with simulation results. My algorithm implementation uses communication messages of fixed size and constant memory on each robot, and is a practical solution for large populations of low-cost robots.
Synthesis for Stochastic Robotic Systems
(2021-08-13) Wells, Andrew Marshall; Kavraki, Lydia E; Vardi, Moshe Y.
Robots interact with their environment, other robots or humans. We need ways to guarantee safety and generally “correct” behavior. This requires a way to specify correct behavior and a model of the human-environment system. We formulate the problem as a game the robot plays against the environment. In this work, we discuss a set of approaches to increasing robot reliability. Essentially, all of these approaches consider models of robot and environment actions and compute potential problems that may arise as well as solutions. This can allow the robot to avoid scenarios where a problem would inevitably lead to some disaster, assuming some better alternative exists. There are essentially two stages to this problem. In the first stage, we model the robot and environment so that we can mathematically reason about potential scenarios and the probabilities of various outcomes. In the second stage we “solve” the model by computing a strategy that the robot can follow to optimally achieve its goal.
Towards Robust Planning for High-DoF Robots in Human Environments: The Role of Optimization
(2024-08-09) Quintero Pena, Carlos; Kavraki, Lydia E; Kyrillidis, Anastasios
Robot motion planning has been a key component in the race to achieve true robot autonomy. It encompasses methods to generate robot motion that meets kinematic constraints, robot dynamics and that is safe (avoids colliding with the environment). It has been particularly successful in efficiently finding motions for high degree-of-freedom robots such as manipulators, but despite tremendous advances, motion planning methods are not ready for human environments. The uncertainty, diversity and clutter of the human world challenge the assumptions of motion planning methods breaking their guarantees, rendering them useless or dramatically worsening their performance. In this thesis, we propose methods to address three important challenges in augmenting motion planning and long-horizon manipulation for human environments. First, we present a framework that enables human-guided motion planning and demonstrate how it can be used for safe planning in partially-observed environments. Second, we present two methods for safe motion planning in the presence of sensing uncertainty, one that requires the poses of segmented objects and another one that acts directly on distance information from a noisy sensor. Finally, we present a framework that dramatically improves the performance of long-horizon manipulation tasks in the presence of clutter for an important class of manipulation problems. All of our contributions have mathematical optimization as a connecting thread to synthesize high-dimensional trajectories using low-dimensional information or as a layer between high-level and low-level planners. Our results demonstrate how these formulations can be effectively used to augment motion planning and planning for manipulation in novel ways, attaining more robust, efficient and reliable methods.

Browsing by Author "Kavraki, Lydia E"

Results Per Page

Sort Options