Generating Better Conformations for Roadmaps in Protein Folding

Size: px
Start display at page:

Download "Generating Better Conformations for Roadmaps in Protein Folding"

Transcription

1 Generating Better Conformations for Roadmaps in Protein Folding Jeff May, Lydia Tapia, and Nancy M. Amato Parasol Lab, Dept. of Computer Science, Texas A&M University, College Station, TX {ltapia, Abstract Probabilistic roadmap methods (PRM) in robotics have been successfully applied to find accurate protein folding pathways with a fast and robust framework for proteins with around 6 3 residues. We can create a roadmap that can be queried multiple times for energy feasible pathways to particular folded states. Misfolded proteins are the cause of proteopathic diseases and are of interest when studying folding pathways. For example, Alzheimer s disease, which is caused by the misfolding of Amyloid Precursor Protein, has between 365 to 77 amino acids[5]. Our goal is to generate better quality roadmaps for larger proteins. For the node generation phase, we generate new protein conformations from old conformations by perturbing them at random points along the protein chain. Certain portions of the protein contain bonds in the folded state that would restrict movement of the protein. Rigidity analysis allows us to classify these residues of a protein, and separate the protein, into two regions: rigid and flexible. We can specify different probabilities of perturbing the conformation for each region to improve the node generation process[2]. In this paper focus on improving the node generation process for PRMs. First, we look at the effect that these parameter values have on the resulting conformations. With this information we construct a Markov Decision Process (MDP) policy learning algorithm to dynamically tune parameter values for the node generation process. We show that policy learning shows promise to automate the task of specifying good parameters.

2 2 Introduction Using graphs to represent protein folding pathways is quick and robust method for finding multiple pathways in a very short amount of time using a graph traversal method. To generate the roadmap for protein folding, we borrow the probabilistic roadmap method (PRM) from robotics and model the protein as we would a robot with many armed linkages fixed to each other. Our goal is to find pathways that are likely to represent the way in which proteins will fold to reach particular conformations. Instead of having our robot maneuver around obstacles, it moves through a funnel-like potential energy landscape along energy feasible paths. PRMs have been implemented for a wide variety of applications[6]. For this research, we focus on the roadmap construction phase, specifically, the node generation process. We generate a new node by perturbing one or more of the angles of a parent conformation from the roadmap. If it is energy feasible, we add it to the roadmap. We start the process from the native state, the conformation representing the folded and functional protein in its natural environment. Rigidity analysis is a simple and fast method for classifying regions of the protein. We can use different probabilities of perturbing the protein in these regions to improve the node generation process. The issue then becomes picking the right parameter values for this process. By using a Markov Decision Process to reward the use of good parameters, we can dynamically tune the parameters during the node generation process produce nodes within an area of interest. For our process, we divide the energy landscape funnel into bins and attempt to fill each bin to a desired number of nodes. The process requires finding a sufficient number of nodes in each layer. By adapting the selection of parameters with MDP Policy Learning, we automatically find useful parameters for different areas of the funnel. This will save researchers valuable time for selecting good parameters on different varieties of proteins. 3 Related Work 3. Modeling Protein Folding with PRMs A probabilistic roadmap method (PRM) is a technique from robotics that is used for planning the motion of a robot from a start position to a goal position. It relies on the ability to abstract all the possible ways a robot can be positioned into what is called configuration space. The number of degrees of freedom (DoFs) that is required to describe the exact position of the robot determines the dimensionality of its configuration space. Mapping this space completely becomes an intractable problem very quickly. Instead we can use probabilistic methods, such as PRMs, to solve these problems with probabilistic completeness, where as the number of samples of configuration space increases, the probability of completely covering the space approaches %. We model a protein conformation as a chain of linkages, representing the backbone position of the protein[9]. Each residue has two torsional directions of rotation. A conformation, in our model, is a set of all the angles between the backbones of the amino acids/residues in the protein. Since proteins have 2 torsional degrees of freedom between each pair of amino acids, they have twice as many DoFs as they have amino acids. The set of all possible conformations (C-space) has as many dimensions as the protein has degrees of freedom. This makes the task of covering the landscape difficult for even small proteins, and it rapidly becomes more difficult for larger proteins. PRMs have two phases: a construction phase and a query phase. A roadmap is basically a simple graph containing valid configurations of a robot and collision-free connections between them. Once this graph is obtained, it can solve many queries for pathways in a short amount of time. The challenge of PRMs, and the focus of this research, is to generate good quality roadmaps. The quality of a roadmap depends on its use. For protein folding, we are interested in mapping the potential energy landscape to provide energy-feasible pathways between various protein conformations. A good quality roadmap will have well placed nodes that are well connected and cover the features of the energy landscape. 3.2 Roadmap Construction for Proteins To create a roadmap, we sample various points in C-space. Our methods bias sampling to increase density of nodes near the native state. The potential energy of these conformations / nodes are calculated. We use this information to decide whether to add the node to the roadmap or not. Once we have a set of nodes, we connect the nodes and test whether the transition between each of the connected conformations are energy feasible, discarding those that are not.

3 3.3 Node Generation for Better Protein Conformations We sample the configuration space of the protein by storing nodes that represent particular conformations. We classify regions of these conformations using rigidity analysis[2]. Each residue in the conformation is classified as part of a rigid, flexible, or dependently flexible region. A rigid region is one in which the bonds on the molecule would prevent the movement of the residues in any direction. A flexible region is defined by the characteristic that the residues are not held by bonds and are not constricted from changing their angle, and where doing so would not necessitate that the rest of the protein change. A dependently flexible region is one in which there is flexibility in a region, however changing the angle in any of the residues would require the rest of the protein to move. We generate nodes incrementally by randomly perturbing a subset of the nodes we currently have in the roadmap. To perturb a given conformation, we move along the amino acid chain until we decide to bend the angle of a particular residue using the following parameters: P flex the probability of perturbing a flexible region P rigid the probability of perturbing a rigid region θ std the angle (or quantity) of perturbation This method works better than the other methods and requires a smaller roadmap to compute the correct formation orders[2]. However, this method introduces a new problem: choosing the right parameter values. Choosing between many possible permutations of parameter values can be cumbersome for a researcher. A machine learning process for selecting good parameter values would ease the process of choosing the right parameter values. 3.4 Rigidity Analysis We use a rigidity analysis technique belonging to the class of approaches called the pebble game[, ] to better simulate motion. It is fast and efficient; we can apply it to every conformation we sample. The pebble game is a constraint counting algorithm which determines the DoF in a two-dimensional graph, along with its rigid/flexible regions. In 2D, the pebble game assigns each vertex two pebbles, representing its two DoF, see Figure a. Each edge/constraint is examined to determine if it is independent or redundant. If two free pebbles can be placed on both endpoints of the edge, then it is marked independent and covered by a pebble from one of its incident vertices. Once an edge is covered by a pebble, it remains covered, although which vertex the pebble comes from may change. Pebbles may be rearranged as shown in Figure a. If pebbles cannot be rearranged to get two free pebbles on both of an edges endpoints, then the edge is marked redundant and indicates a rigid region. In the end, the remaining free pebbles indicate the graphs DoF. The 2D pebble does not generalize to 3D for arbitrary graphs, but it can be applied to 3D bond-bending network[]. A bond-bending network is a truss structure with constraints between nearest neighbors and nextnearest neighbors. A protein, with xed bond lengths and bond angles, can be modeled as a bond-bending network, called the bar-joint model, where atoms are modeled as vertices with 3 DoF and bonds are modeled as edges, see Figure b. It has been successfully used by several applications to study protein rigidity and flexibility[2, 8, 8, 5]. An alternative model, the body-bar model, represents atoms as rigid bodies with 6 DoF and the torsional bonds between them as 5 bars/constraints[22] see Figure c. Both models are conjectured to be equivalent[]. move on move off (a) (b) (c) Figure : (a) The result of the pebble game on a 2D graph. Pebbles may be free (white) or covering (black). Constraints are marked as independent (solid) or redundant (dashed). Pebbles may be rearranged as shown. Rigidity models for a sample molecule: (b) bar-joint and (c) body-bar.

4 3.5 Connecting the Roadmap In the second step, connections (edges) are made between sampled conformations with similar structure. Weights are assigned to directed edges to reflect the energetic feasibility of transitioning between the two endpoint conformations. This combination of nodes and weighted edges forms a roadmap that approximates the energy landscape. This roadmap encodes thousands of folding pathways. The most energetically feasible pathways in the roadmap can be extracted using these weights. Connections between two nodes, q and q 2, are labeled with edge weights that reflect the energetic feasibility of transitioning between them. This is done by first identifying all the intermediate nodes, q = c, c,..., c n, c n = q 2, that connect q to q 2. For each pair of consecutive conformations c i and c i+, the probability P i of transitioning from c i to c i+ depends on the difference between their potential energies E i = E(c i+)e(c i): P i = { e E i kt if E i > if E i () This keeps the detailed balance between two adjacent states and enables the edge weight to be computed by summing the logarithms of the probabilities for all pairs of consecutive conformations in the sequence. With this edge weight definition, we can use simple graph search algorithms to extract the most energetically feasible pathways in the roadmap between two given states (e.g. from the unfolded state to the folded state). 3.6 Potential Energy Calculation Our method is flexible and allows any potential function to be used. In this paper, we use a coarse potential function similar to [6]. We use a step function approximation of the van der Waals potential component and model side chains as spheres with zero DoF. If any two spheres are too close (i.e., less than 2.4Å during sampling and.å during connection), a very high potential is returned. Otherwise, the potential is: U tot = K d {[(d i d ) 2 + d 2 c] /2 d c} + E hp (2) restraints where K d is kcal/mol and d = d c = 2. The rst term represents constraints favoring known secondary structure through main-chain hydrogen bonds and disulphide bonds, and the second term is the hydrophobic effect. The hydrophobic eect is computed as follows: if two hydrophobic residues are within 6 of each other, then the potential is decreased by kj/mol. In our previous work [2, 2], we provided methods for building an approximate map of a proteins potential energy landscape[2, 2] and an RNAs folding landscape[2]. We have published results from our approximate maps for proteins up to 48 residues easily built on a desktop PC[2]. Our roadmaps give an approximate view of the protein folding landscape. In the past, we have successfully extracted low-energy pathways, validated secondary structure formation order, and seen general and consistent trends in reaction coordinates such as native contacts present and RMSD. 3.7 Energy feasibility We use the potential energy calculation from 3.6 to decide the probability of accepting a conformation into the roadmap. A conformation q, with potential energy E(q), is considered energy feasible with the probability: P(accept q) = if E(q) < E min if E min E(q) E max (3) if E(q) > E max E max E(q) E max E min where E min is the potential energy of the open chain and E max is 2E min. 3.8 Native contacts present A native contact is a pair of C α atoms that are within a specified range in a given conformation as well as in the native state. For our system, we use at a distance of less than 7Å to define whether a pair of C α atoms are in contact. The number of native contacts present in any conformation can be used as an indication of how close it is to the native state. These contacts are used to define the formation of secondary structures in a conformation and to calculate the formation orders in a roadmap.

5 4 Methodology 4. Studying the Parameters The first stage of our research was to study the effect that different parameter values have on the resulting conformations. We collected data on the resulting conformations, such as the number of native contacts, the potential energy, and whether the node was energy feasible or in collision. When building a model of the energy landscape, we have two primary goals. First, we expect good coverage of the energy landscape. We would like a set of conformations that represent the allowable motions of the molecule. Second, we would like high-quality conformations. A high-quality conformation is one that will be feasibly undertaken, e.g., low energy. 4.2 MDP Policy Learning for Selecting Parameter Sets 4.2. Markov Decision Process in PRMs MDP policy learning has been previously used to impact PRM sampling. In robotics applications, there are many proposed sampling methods[3,, 4, 4, 7, 3] whose efficiency and effectiveness has been seen to be highly correlated with the planning space and the problem construction[7]. A method called Hybrid PRM uses MDP policy learning with a library of possible sampling methods, the actions. It has been shown to automatically learn which sampling methods will best cover an unclassified problem[9]. Markov decision processes (MDP) occur when an autonomous agent can sense outcomes from its actions in the environment. From the action and outcome relationship, policies can be learned to choose optional actions to achieve the agent s goals. Markov decision processes occur in many domains and has been solved in problems including mobile robot control, game playing, and robot motion planning. In a MDP, the learning agent perceives a set of states, S, that describe its current environment. It also has a set of actions, A, that it can select from. At each time step t, it can select an action at that is taken. The outcome of that action makes some impact on the environment β(s t, a t) that results in a new set of perceptions, s t+. The result of that outcome is measured in a reward function, r(s t, a t), that gives the agent some knowledge of the utility of its action a t. Learning progresses by the agent s pursuit of maximal rewards. Solutions to MDPs are commonly solved through dynamic programming and reinforcement learning. Algorithm 4. MDP Policy Learning in Node Generation Input. A set actions A defined by the parameters that can be selected a set of states S that defines current perception, a starting conformation c : for timestep, t, from to n do 2: Select action a t from probabilistically from rankings R a of set A 3: Generate c t where c t β(a t, c t ) 4: Generate reward r t based on new perceptions, s t+ where s t+ δ(s t, a t ) 5: Apply reward r t to P a where a is a t 6: end for Output. A set rankings of R a that defines the utility of each A Reward Function for Layers Method In order to insure that we have a diverse spread of nodes, we store conformations with similar numbers of native contacts into a common bin. Each bin holds conformations within a certain range of native contacts. We reward parameter sets for producing a diverse conformation. For each bin, we attempt to fill it to a desired amount. Rewards are generated for Algorithm 4. as follows based on the resulting conformation: R max: the conformation is within the bin we are attempting to fill R partial : the conformation is within another bin that is not currently filled R penalty : the conformation has an energy greater than E max or it is in a layer that is already full

6 4.2.3 Choosing a Parameter Set To choose a set of parameters for node generation (see Algorithm 4.), we use the roulette wheel selection method. This method is analogous to having a slot for every permutation of the possible parameter values for P flex, P rigid, θ flex, and θ rigid on a roulette wheel, where the size of the slot P i is dependent on its score r i relative to the scores of all the other parameter sets R. It probabilistically selects a set of parameters a from A by performing a random walk along the circumference of the wheel. Every permutation starts with an equal probability of being chosen. Once it has received a reward, it adds it to its score. Algorithm 4.2 Selecting a Parameter Set in Policy Learning Input. A set of all possible parameter set permutations A, a set of scores R for each value in A initially set to, a learning rate L : if random value v [, ) < L then 2: return a random parameter set, a A 3: else 4: Select a random velocity, v, such that: v < ( A + R) 5: if v < A then 6: return a random parameter set, a A 7: else 8: i 9: while v > and r i R do : v vr i : i i + 2: end while 3: return a i 4: end if 5: end if Output. a parameter set, a A As we add scores, the probability of selecting a particular parameter set changes. Therefore, at a given stage in Algorithm 4.2, the probability of selecting the parameter set a i is given by: ( n ) P i = ( + R i) / R i, where n = R (4) 5 Results 5. Parameter Study To test the effect the parameters had on the resulting nodes, we generated maps for proteins G, BDD, and 2CRS, each with 4 nodes. We ran the tests by randomly selecting parameter set from the value sets, {,,,,, } for P flex and P rigid and {.27,.55,.83,.38,.277,.555,.} for θ std in radians. We collected the potential energy, number of native contacts, and whether it was energy feasible. By plotting average rate of energy feasible nodes generated per attempt for every parameter set combination we attain Figure 2. In this figure, the spacial dimensions (x, y,z) correspond to the value of the parameter in the set (P flex, P rigid, θ std ). We use a fourth dimension of color to represent the energy feasibility rate, where a dark color represents a high energy feasibility rate, since a smaller proportion of nodes are discarded, and a lighter color represents low energy feasibility, thus a greater proportion is discarded. In Figure 3, we plot the average number of native contacts that each parameter combination produces. Both Figure 2 and 3 were generated from the same test on protein G (PDB ID: GB) using 4, nodes. If the parameter values varied independently from each other, we would see that the colors along that slice of the 3 dimensional space would remain consistent. In Figure 2, we see that the color varies dramatically, this demonstrates that they are greatly dependent on each other. Although the parameter values behave similar to our expectations, specifying the right parameters is more difficult than choosing the values independent of each other.

7 Proportion of Energy Feasible Nodes stdangle prigid pflex Figure 2: Effect of the parameters P rigid, P flex, and θ std on the rate of energy feasible generated for Protein G (PDB ID GB). Average Native Contacts.75.7 stdangle prigid pflex Figure 3: Effect of the parameters P rigid, P flex, and θ std on the average number of native contacts of the generated nodes for Protein G (PDB ID GB). 5.2 Policy Learning After implementing the policy learning, we examined the number of times a parameter set was chosen. We used a set {,,, } for the P flex and P rigid values, a subset of the values from the previous tests. These tests were run with R max =, R partial = 5, and R penalty = -.2. We used a learning rate L of.5. After filling half of the bins, the scores are reset: R. However, in these tests, we used different angle values for flexible and rigid regions, θ flex and θ rigid. For Figure 4, we converted these θ values to a single angle value θ std = P flex θ flex +P rigid θ rigid. We would expect that the parameter values would match the energy feasibility from Figure 2, with the added prospect of finding diverse conformations. MDP policy learning has been previously used to impact PRM sampling. The results in Figure 4 show that the algorithm was learning to use parameter sets for reasons other than just energy feasibility. By examining the color at P flex =, P rigid =, and the θ std is greatest, we see that this parameter set was chosen many times, despite having a lower energy feasibility, this is caused by the parameter set being rewarded for generating diverse nodes.

8 Number of Selections stdangle prigid pflex Figure 4: The number of times the parameter set of {P rigid, P flex, θ std } was chosen using policy learning with a learning rate of.5 on Protein G (PDB ID GB). 6 Conclusions and Future Work The preliminary tests showed that the outcome of selecting a particular parameter value depended strongly on the other parameter values selected (Figure 2). This makes the problem of specifying good parameter values before running each simulation difficult. In order to alleviate the researcher from having to specify only good parameter values, we can automatically learn which of the values specified are performing well, and bias the selection process in favor of those values. The policy learning method dynamically adapts to the outcomes of the node generation process. By examining the values that were selected, we can find what parameter sets the algorithm found to work best at that time. These values are not always obvious and can vary depending on which conformations have been generated. By specifying a wide range of values, policy learning can hone in on the useful ones. Since we have to specify a learning rate along with our parameter, we add another parameter to specify. For future work, we could study the effect the learning rate has for different sized and shaped proteins. We could also test other learning algorithms. However, policy learning is quick and simple method for choosing actions, and we have shown that it can be used to ease the process of selecting good parameter values. We can also test this algorithm on larger proteins to see if it is capable of adapting and generating better quality maps in shorter amounts of time. With enough improvement, we could hopefully study larger proteins to find folding pathways for proteopathic diseases, such as Alzhiemer s or Mad Cow disease. We have a publicly available protein folding and motion server at: Given a single protein conformation/pdb file, our server will simulate the folding behavior and provide several different views of extracted folding pathways. Given multiple protein conformations/pdb files, our server will simulate the transition pathways between the different input states. We invite the community to help enrich our web-based motion database by submitting PDB files for study. References [] N. M. Amato, O. B. Bayazit, L. K. Dale, C. V. Jones, and D. Vallejo. OBPRM: An obstacle-based PRM for 3D workspaces. In Robotics: The Algorithmic Perspective, pages 55 68, Natick, MA, 998. A.K. Peters. Proc. Third Workshop on Algorithmic Foundations of Robotics (WAFR), Houston, TX, 998. [2] N. M. Amato, K. A. Dill, and G. Song. Using motion planning to map protein folding landscapes and analyze folding kinetics of known native structures. J. Comput. Biol., (3-4): , 23. Special issue of Int. Conf. Comput. Molecular Biology (RECOMB) 22. [3] R. Bohlin and L. E. Kavraki. Path planning using Lazy PRM. In Proc. IEEE Int. Conf. Robot. Autom. (ICRA), pages , 2. [4] V. Boor, M. H. Overmars, and A. F. van der Stappen. The Gaussian sampling strategy for probabilistic roadmap planners. In Proc. IEEE Int. Conf. Robot. Autom. (ICRA), volume 2, pages 8 23, May 999.

9 [5] F. Chiti and C. Dobson. Protein misfolding, functional amyloid, and human disease. Annu. Rev. Biochem., 75: , 26. [6] H. Choset, K. M. Lynch, S. Hutchinson, G. A. Kantor, W. Burgard, L. E. Kavraki, and S. Thrun. Principles of Robot Motion: Theory, Algorithms, and Implementations. MIT Press, Cambridge, MA, June 25. [7] R. Geraerts and M. H. Overmars. Reachablility-based analysis for probabilistic roadmap planners. Robotics and Autonomous Systems, 55: , 27. [8] B. M. Hespenheide, A. Rader, M. Thorpe, and L. A. Kuhn. Identifying protein folding cores from the evolution of flexible regious during unfolding. J. Mol. Gra. Model., 2:95 27, 22. [9] D. Hsu, G. Sánchez-Ante, and Z. Sun. Hybrid PRM sampling with a cost-sensitive adaptive strategy. In Proc. IEEE Int. Conf. Robot. Autom. (ICRA), pages , 25. [] D. Jacobs. Generic rigidity in three-dimensional bond-bending networks. J. Phys. A: Math. Gen., 3: , 998. [] D. Jacobs and M. Thorpe. Generic rigidity percolation: The pebble game. Phys. Rev. Lett., 75(22):45 454, 995. [2] D. J. Jacobs, A. Rader, L. A. Kuhn, and M. Thorpe. Protein flexiblility predictions using graph theory. Proteins Struct. Funct. Genet., 44:5 65, 2. [3] L. E. Kavraki, P. Švestka, J. C. Latombe, and M. H. Overmars. Probabilistic roadmaps for path planning in high-dimensional configuration spaces. IEEE Trans. Robot. Automat., 2(4):566 58, August 996. [4] S. M. LaValle and J. J. Kuffner. Randomized kinodynamic planning. In Proc. IEEE Int. Conf. Robot. Autom. (ICRA), pages , 999. [5] M. Lei, M. I. Zavodszky, L. A. Kuhn, and M. F. Thorpe. Sampling protein conformations and pathways. J. Comput. Chem., 25:33 48, 24. [6] M. Levitt. Protein folding by restrained energy minimization and molecular dynamics. J. Mol. Biol., 7: , 983. [7] C. L. Nielsen and L. E. Kavraki. A two level fuzzy PRM for manipulation planning. Technical Report TR2-365, Computer Science, Rice University, Houston, TX, 2. [8] A. Rader, B. M. Hespenheide, L. A. Kuhn, and M. Thorpe. Protein unfolding: Rigidity lost. Proc. Natl. Acad. Sci. USA, 99(6): , 22. [9] M. J. Sternberg. Protein Structure Prediction. OIRL Press at Oxford University Press, 996. [2] X. Tang, S. Thomas, L. Tapia, and N. M. Amato. Tools for simulating and analyzing RNA folding kinetics. In Proc. Int. Conf. Comput. Molecular Biology (RECOMB), pages , 27. [2] S. Thomas, X. Tang, L. Tapia, and N. M. Amato. Simulating protein motions with rigidity analysis. In Proc. Int. Conf. Comput. Molecular Biology (RECOMB), pages , 26. [22] W. Whiteley. Some matroids from discrete applied geometry. Contemp. Math., 97:7 3, 996.

Predicting Protein Folding Paths. S.Will, , Fall 2011

Predicting Protein Folding Paths. S.Will, , Fall 2011 Predicting Protein Folding Paths Protein Folding by Robotics Probabilistic Roadmap Planning (PRM): Thomas, Song, Amato. Protein folding by motion planning. Phys. Biol., 2005 Aims Find good quality folding

More information

vizmo++: a Visualization, Authoring, and Educational Tool for Motion Planning

vizmo++: a Visualization, Authoring, and Educational Tool for Motion Planning vizmo++: a Visualization, Authoring, and Educational Tool for Motion Planning Aimée Vargas E. Jyh-Ming Lien Nancy M. Amato. aimee@cs.tamu.edu neilien@cs.tamu.edu amato@cs.tamu.edu Technical Report TR05-011

More information

Adaptive Neighbor Connection Aids Protein Motion Modeling

Adaptive Neighbor Connection Aids Protein Motion Modeling Adaptive Neighbor Connection Aids Protein Motion Modeling Chinwe Ekenna,Shawna Thomas,Nancy M. Amato Parasol Lab, Department of Computer Science and Engineering, Texas A&M University, College Station,

More information

Exploiting collision information in probabilistic roadmap planning

Exploiting collision information in probabilistic roadmap planning Exploiting collision information in probabilistic roadmap planning Serene W. H. Wong and Michael Jenkin Department of Computer Science and Engineering York University, 4700 Keele Street Toronto, Ontario,

More information

Probabilistic Motion Planning: Algorithms and Applications

Probabilistic Motion Planning: Algorithms and Applications Probabilistic Motion Planning: Algorithms and Applications Jyh-Ming Lien Department of Computer Science George Mason University Motion Planning in continuous spaces (Basic) Motion Planning (in a nutshell):

More information

1 Introduction Automatic motion planning deals with finding a feasible sequence of motions to take some moving object (the `robot') from a given initi

1 Introduction Automatic motion planning deals with finding a feasible sequence of motions to take some moving object (the `robot') from a given initi Customizing PRM Roadmaps at Query Time Λ Guang Song Shawna Miller Nancy M. Amato Dept. of Computer Science Dept. of Computer Science Dept. of Computer Science gsong@cs.tamu.edu slm5563@cs.tamu.edu amato@cs.tamu.edu

More information

Studying Protein Folding Using Motion Planning Techniques

Studying Protein Folding Using Motion Planning Techniques Studying Protein Folding Using Motion Planning Techniques Susan S. Lin sslin@eecs.berkeley.edu Dr. Nancy M. Amato, Faculty Advisor Guang Song, Grad Student Advisor amato,gsong @cs.tamu.edu Department of

More information

UOBPRM: A Uniformly Distributed Obstacle-Based PRM

UOBPRM: A Uniformly Distributed Obstacle-Based PRM UOBPRM: A Uniformly Distributed Obstacle-Based PRM Hsin-Yi (Cindy) Yeh 1, Shawna Thomas 1, David Eppstein 2 and Nancy M. Amato 1 Abstract This paper presents a new sampling method for motion planning that

More information

Incremental Map Generation (IMG)

Incremental Map Generation (IMG) Incremental Map Generation (IMG) Dawen Xie, Marco Morales, Roger Pearce, Shawna Thomas, Jyh-Ming Lien, and Nancy M. Amato Parasol Lab, Department of Computer Science, Texas A&M University, College Station,

More information

II. RELATED WORK. A. Probabilistic roadmap path planner

II. RELATED WORK. A. Probabilistic roadmap path planner Gaussian PRM Samplers for Dynamic Configuration Spaces Yu-Te Lin and Shih-Chia Cheng Computer Science Department Stanford University Stanford, CA 94305, USA {yutelin, sccheng}@cs.stanford.edu SUID: 05371954,

More information

Incremental Map Generation (IMG)

Incremental Map Generation (IMG) Incremental Map Generation (IMG) Dawen Xie, Marco A. Morales A., Roger Pearce, Shawna Thomas, Jyh-Ming Lien, Nancy M. Amato {dawenx,marcom,rap2317,sthomas,neilien,amato}@cs.tamu.edu Technical Report TR06-005

More information

RESAMPL: A Region-Sensitive Adaptive Motion Planner

RESAMPL: A Region-Sensitive Adaptive Motion Planner REAMPL: A Region-ensitive Adaptive Motion Planner amuel Rodriguez, hawna Thomas, Roger Pearce, and Nancy M. Amato Parasol Lab, Department of Computer cience, Texas A&M University, College tation, TX UA

More information

Providing Haptic Hints to Automatic Motion Planners

Providing Haptic Hints to Automatic Motion Planners Providing Haptic Hints to Automatic Motion Planners O. Burchan Bayazit Guang Song Nancy M. Amato fburchanb,gsong,amatog@cs.tamu.edu Department of Computer Science, Texas A&M University College Station,

More information

Customizing PRM Roadmaps at Query Time Λ

Customizing PRM Roadmaps at Query Time Λ Customizing PRM Roadmaps at Query Time Λ Guang Song Shawna Miller Nancy M. Amato Department of Computer Science Texas A&M University College Station, TX 77843-3112 fgsong,slm5563,amatog@cs.tamu.edu Abstract

More information

Probabilistic Roadmap Planner with Adaptive Sampling Based on Clustering

Probabilistic Roadmap Planner with Adaptive Sampling Based on Clustering Proceedings of the 2nd International Conference of Control, Dynamic Systems, and Robotics Ottawa, Ontario, Canada, May 7-8 2015 Paper No. 173 Probabilistic Roadmap Planner with Adaptive Sampling Based

More information

Planning with Reachable Distances: Fast Enforcement of Closure Constraints

Planning with Reachable Distances: Fast Enforcement of Closure Constraints Planning with Reachable Distances: Fast Enforcement of Closure Constraints Xinyu Tang Shawna Thomas Nancy M. Amato xinyut@cs.tamu.edu sthomas@cs.tamu.edu amato@cs.tamu.edu Technical Report TR6-8 Parasol

More information

Motion Planning with Dynamics, Physics based Simulations, and Linear Temporal Objectives. Erion Plaku

Motion Planning with Dynamics, Physics based Simulations, and Linear Temporal Objectives. Erion Plaku Motion Planning with Dynamics, Physics based Simulations, and Linear Temporal Objectives Erion Plaku Laboratory for Computational Sensing and Robotics Johns Hopkins University Frontiers of Planning The

More information

Workspace Skeleton Tools for Motion Planning

Workspace Skeleton Tools for Motion Planning Workspace Skeleton Tools for Motion Planning Kiffany Lyons, Diego Ruvalcaba, Mukulika Ghosh, Shawna Thomas, and Nancy Amato Abstract Motion planning is the ability to find a valid path from a start to

More information

OOPS for Motion Planning: An Online, Open-source, Programming System

OOPS for Motion Planning: An Online, Open-source, Programming System IEEE Inter Conf on Robotics and Automation (ICRA), Rome, Italy, 2007, pp. 3711 3716 OOPS for Motion Planning: An Online, Open-source, Programming System Erion Plaku Kostas E. Bekris Lydia E. Kavraki Abstract

More information

1 Introduction Folding is a very common process in our lives, ranging from the macroscopic level paper folding or gift wrapping to the microscopic lev

1 Introduction Folding is a very common process in our lives, ranging from the macroscopic level paper folding or gift wrapping to the microscopic lev A Motion Planning Approach to Folding: From Paper Craft to Protein Folding Λ Guang Song gsong@cs.tamu.edu Nancy M. Amato amato@cs.tamu.edu Technical Report TR00-017 Department of Computer Science Texas

More information

Workspace Skeleton Tools for Motion Planning

Workspace Skeleton Tools for Motion Planning Workspace Skeleton Tools for Motion Planning Diego Ruvalcaba, Kiffany Lyons, Mukulika Ghosh, Shawna Thomas, and Nancy Amato Abstract Motion planning is the ability to find a valid path from a start to

More information

Parallel Protein Folding with STAPL

Parallel Protein Folding with STAPL Parallel Protein Folding with STAPL Shawna Thomas Nancy M. Amato Dept. of Computer Science Teas A&M University {sthomas, amato}@cs.tamu.edu Abstract The protein folding problem is to study how a protein

More information

Workspace Importance Sampling for Probabilistic Roadmap Planning

Workspace Importance Sampling for Probabilistic Roadmap Planning Workspace Importance Sampling for Probabilistic Roadmap Planning Hanna Kurniawati David Hsu Department of Computer Science National University of Singapore Singapore, Republic of Singapore {hannakur, dyhsu}@comp.nus.edu.sg

More information

Sampling-Based Robot Motion Planning. Lydia Kavraki Department of Computer Science Rice University Houston, TX USA

Sampling-Based Robot Motion Planning. Lydia Kavraki Department of Computer Science Rice University Houston, TX USA Sampling-Based Robot Motion Planning Lydia Kavraki Department of Computer Science Rice University Houston, TX USA Motion planning: classical setting Go from Start to Goal without collisions and while respecting

More information

biasing the sampling based on the geometric characteristics of configurations known to be reachable from the assembled configuration (the start). In p

biasing the sampling based on the geometric characteristics of configurations known to be reachable from the assembled configuration (the start). In p Disassembly Sequencing Using a Motion Planning Approach Λ Sujay Sundaram Department of Mechanical Engineering Texas A&M University College Station, TX 77843 sps7711@cs.tamu.edu Ian Remmler Nancy M. Amato

More information

Hybrid Probabilistic RoadMap - Monte Carlo Motion Planning for Closed Chain Systems with Spherical Joints

Hybrid Probabilistic RoadMap - Monte Carlo Motion Planning for Closed Chain Systems with Spherical Joints Hybrid Probabilistic RoadMap - Monte Carlo Motion Planning for Closed Chain Systems with Spherical Joints Li Han Dept. of Mathematics and Computer Science Clark University Worcester, MA 01610 Email: lhan@clarku.edu

More information

A Machine Learning Approach for Feature-Sensitive Motion Planning

A Machine Learning Approach for Feature-Sensitive Motion Planning A Machine Learning Approach for Feature-Sensitive Motion Planning Marco Morales, Lydia Tapia, Roger Pearce, Samuel Rodriguez, and Nancy M. Amato Parasol Laboratory, Dept. of Computer Science, Texas A&M

More information

KINARI-Lib A library for Combinatorial Rigidity analysis and applications

KINARI-Lib A library for Combinatorial Rigidity analysis and applications KINARI-Lib A library for Combinatorial Rigidity analysis and applications Naomi Fox Filip Jagodzinski Ileana Streinu Linkage Lab http://linkage.cs.umass.edu Department of Computer Science Smith College

More information

The Toggle Local Planner for Sampling-Based Motion Planning

The Toggle Local Planner for Sampling-Based Motion Planning The Toggle Local Planner for Sampling-Based Motion Planning Jory Denny and Nancy M. Amato Abstract Sampling-based solutions to the motion planning problem, such as the probabilistic roadmap method (PRM),

More information

Configuration Space of a Robot

Configuration Space of a Robot Robot Path Planning Overview: 1. Visibility Graphs 2. Voronoi Graphs 3. Potential Fields 4. Sampling-Based Planners PRM: Probabilistic Roadmap Methods RRTs: Rapidly-exploring Random Trees Configuration

More information

CS 4649/7649 Robot Intelligence: Planning

CS 4649/7649 Robot Intelligence: Planning CS 4649/7649 Robot Intelligence: Planning Probabilistic Roadmaps Sungmoon Joo School of Interactive Computing College of Computing Georgia Institute of Technology S. Joo (sungmoon.joo@cc.gatech.edu) 1

More information

Deformable Robot Motion Planning in a Reduced-Dimension Configuration Space

Deformable Robot Motion Planning in a Reduced-Dimension Configuration Space 2010 IEEE International Conference on Robotics and Automation Anchorage Convention District May 3-8, 2010, Anchorage, Alaska, USA Deformable Robot Motion Planning in a Reduced-Dimension Configuration Space

More information

Rapidly-Exploring Roadmaps: Weighing Exploration vs. Refinement in Optimal Motion Planning

Rapidly-Exploring Roadmaps: Weighing Exploration vs. Refinement in Optimal Motion Planning Rapidly-Exploring Roadmaps: Weighing Exploration vs. Refinement in Optimal Motion Planning Ron Alterovitz, Sachin Patil, and Anna Derbakova Abstract Computing globally optimal motion plans requires exploring

More information

A Hybrid Approach for Complete Motion Planning

A Hybrid Approach for Complete Motion Planning A Hybrid Approach for Complete Motion Planning Liangjun Zhang 1 Young J. Kim 2 Dinesh Manocha 1 1 Dept. of Computer Science, University of North Carolina at Chapel Hill, USA, {zlj,dm}@cs.unc.edu 2 Dept.

More information

Geometric Path Planning McGill COMP 765 Oct 12 th, 2017

Geometric Path Planning McGill COMP 765 Oct 12 th, 2017 Geometric Path Planning McGill COMP 765 Oct 12 th, 2017 The Motion Planning Problem Intuition: Find a safe path/trajectory from start to goal More precisely: A path is a series of robot configurations

More information

Probabilistic Roadmaps of Trees for Parallel Computation of Multiple Query Roadmaps

Probabilistic Roadmaps of Trees for Parallel Computation of Multiple Query Roadmaps Probabilistic Roadmaps of Trees for Parallel Computation of Multiple Query Roadmaps Mert Akinc, Kostas E. Bekris, Brian Y. Chen, Andrew M. Ladd, Erion Plaku, and Lydia E. Kavraki Rice University Department

More information

Workspace-Guided Rapidly-Exploring Random Tree Method for a Robot Arm

Workspace-Guided Rapidly-Exploring Random Tree Method for a Robot Arm WorkspaceGuided RapidlyExploring Random Tree Method for a Robot Arm Jaesik Choi choi31@cs.uiuc.edu August 12, 2007 Abstract Motion planning for robotic arms is important for real, physical world applications.

More information

Learning to Guide Random Tree Planners in High Dimensional Spaces

Learning to Guide Random Tree Planners in High Dimensional Spaces Learning to Guide Random Tree Planners in High Dimensional Spaces Jörg Röwekämper Gian Diego Tipaldi Wolfram Burgard Fig. 1. Example paths for a mobile manipulation platform computed with RRT-Connect [13]

More information

Figure 1: A typical industrial scene with over 4000 obstacles. not cause collisions) into a number of cells. Motion is than planned through these cell

Figure 1: A typical industrial scene with over 4000 obstacles. not cause collisions) into a number of cells. Motion is than planned through these cell Recent Developments in Motion Planning Λ Mark H. Overmars Institute of Information and Computing Sciences, Utrecht University, P.O. Box 80.089, 3508 TB Utrecht, the Netherlands. Email: markov@cs.uu.nl.

More information

A Hybrid Approach for Complete Motion Planning

A Hybrid Approach for Complete Motion Planning A Hybrid Approach for Complete Motion Planning Liangjun Zhang 1 Young J. Kim 2 Dinesh Manocha 1 1 Dept. of Computer Science, University of North Carolina at Chapel Hill, USA, {zlj,dm}@cs.unc.edu 2 Dept.

More information

6.141: Robotics systems and science Lecture 9: Configuration Space and Motion Planning

6.141: Robotics systems and science Lecture 9: Configuration Space and Motion Planning 6.141: Robotics systems and science Lecture 9: Configuration Space and Motion Planning Lecture Notes Prepared by Daniela Rus EECS/MIT Spring 2012 Figures by Nancy Amato, Rodney Brooks, Vijay Kumar Reading:

More information

Sampling-based Planning 2

Sampling-based Planning 2 RBE MOTION PLANNING Sampling-based Planning 2 Jane Li Assistant Professor Mechanical Engineering & Robotics Engineering http://users.wpi.edu/~zli11 Problem with KD-tree RBE MOTION PLANNING Curse of dimension

More information

Robot Motion Planning

Robot Motion Planning Robot Motion Planning slides by Jan Faigl Department of Computer Science and Engineering Faculty of Electrical Engineering, Czech Technical University in Prague lecture A4M36PAH - Planning and Games Dpt.

More information

Open Access Narrow Passage Watcher for Safe Motion Planning by Using Motion Trend Analysis of C-Obstacles

Open Access Narrow Passage Watcher for Safe Motion Planning by Using Motion Trend Analysis of C-Obstacles Send Orders for Reprints to reprints@benthamscience.ae 106 The Open Automation and Control Systems Journal, 2015, 7, 106-113 Open Access Narrow Passage Watcher for Safe Motion Planning by Using Motion

More information

T S. Configuration Space S1 LP S0 ITM. S3 Start cfg

T S. Configuration Space S1 LP S0 ITM. S3 Start cfg An Adaptive Framework for `Single Shot' Motion Planning Λ Daniel R. Vallejo Christopher Jones Nancy M. Amato Texas A&M University Sandia National Laboratories Texas A&M University dvallejo@cs.tamu.edu

More information

crrt : Planning loosely-coupled motions for multiple mobile robots

crrt : Planning loosely-coupled motions for multiple mobile robots crrt : Planning loosely-coupled motions for multiple mobile robots Jan Rosell and Raúl Suárez Institute of Industrial and Control Engineering (IOC) Universitat Politècnica de Catalunya (UPC) Barcelona

More information

Evaluation of the K-closest Neighbor Selection Strategy for PRM Construction

Evaluation of the K-closest Neighbor Selection Strategy for PRM Construction Evaluation of the K-closest Neighbor Selection Strategy for PRM Construction Troy McMahon, Sam Jacobs, Bryan Boyd, Lydia Tapia, Nancy M. Amato Parasol Laboratory, Dept. of Computer Science and Engineering,

More information

Improving Roadmap Quality through Connected Component Expansion

Improving Roadmap Quality through Connected Component Expansion Improving Roadmap Quality through Connected Component Expansion Juan Burgos, Jory Denny, and Nancy M. Amato Abstract Motion planning is the problem of computing valid paths through an environment. However,

More information

A Motion Planner for a Hybrid Robotic System with Kinodynamic Constraints

A Motion Planner for a Hybrid Robotic System with Kinodynamic Constraints A Motion Planner for a Hybrid Robotic System with Kinodynamic Constraints Erion Plaku Lydia E. Kavraki Moshe Y. Vardi Abstract The rapidly increasing complexity of tasks robotic systems are expected to

More information

Balancing Exploration and Exploitation in Motion Planning

Balancing Exploration and Exploitation in Motion Planning 2008 IEEE International Conference on Robotics and Automation Pasadena, CA, USA, May 19-23, 2008 Balancing Exploration and Exploitation in Motion Planning Markus Rickert Oliver Brock Alois Knoll Robotics

More information

Probabilistic roadmaps for efficient path planning

Probabilistic roadmaps for efficient path planning Probabilistic roadmaps for efficient path planning Dan A. Alcantara March 25, 2007 1 Introduction The problem of finding a collision-free path between points in space has applications across many different

More information

Efficient Planning of Spatially Constrained Robot Using Reachable Distances

Efficient Planning of Spatially Constrained Robot Using Reachable Distances Efficient Planning of Spatially Constrained Robot Using Reachable Distances Xinyu Tang Shawna Thomas Nancy M. Amato xinyut@cs.tamu.edu sthomas@cs.tamu.edu amato@cs.tamu.edu Technical Report TR07-001 Parasol

More information

Useful Cycles in Probabilistic Roadmap Graphs

Useful Cycles in Probabilistic Roadmap Graphs Useful Cycles in Probabilistic Roadmap Graphs Dennis Nieuwenhuisen Mark H. Overmars institute of information and computing sciences, utrecht university technical report UU-CS-24-64 www.cs.uu.nl Useful

More information

A Branch and Bound Algorithm for the Protein Folding Problem in the HP Lattice Model

A Branch and Bound Algorithm for the Protein Folding Problem in the HP Lattice Model Article A Branch and Bound Algorithm for the Protein Folding Problem in the HP Lattice Model Mao Chen* and Wen-Qi Huang School of Computer Science and Technology, Huazhong University of Science and Technology,

More information

Kinodynamic Motion Planning by Interior-Exterior Cell Exploration

Kinodynamic Motion Planning by Interior-Exterior Cell Exploration Kinodynamic Motion Planning by Interior-Exterior Cell Exploration Ioan A. Şucan 1 and Lydia E. Kavraki 1 Department of Computer Science, Rice University, {isucan, kavraki}@rice.edu Abstract: This paper

More information

Local Randomization in Neighbor Selection Improves PRM Roadmap Quality

Local Randomization in Neighbor Selection Improves PRM Roadmap Quality Local Randomization in Neighbor Selection Improves PRM Roadmap Quality Troy McMahon 1, Sam Jacobs 1, Bryan Boyd 1, Lydia Tapia 2, Nancy M. Amato 1 Abstract Probabilistic Roadmap Methods (PRMs) are one

More information

for Motion Planning RSS Lecture 10 Prof. Seth Teller

for Motion Planning RSS Lecture 10 Prof. Seth Teller Configuration Space for Motion Planning RSS Lecture 10 Monday, 8 March 2010 Prof. Seth Teller Siegwart & Nourbahksh S 6.2 (Thanks to Nancy Amato, Rod Brooks, Vijay Kumar, and Daniela Rus for some of the

More information

Clearance Based Path Optimization for Motion Planning

Clearance Based Path Optimization for Motion Planning Clearance Based Path Optimization for Motion Planning Roland Geraerts Mark Overmars institute of information and computing sciences, utrecht university technical report UU-CS-2003-039 www.cs.uu.nl Clearance

More information

Advanced Robotics Path Planning & Navigation

Advanced Robotics Path Planning & Navigation Advanced Robotics Path Planning & Navigation 1 Agenda Motivation Basic Definitions Configuration Space Global Planning Local Planning Obstacle Avoidance ROS Navigation Stack 2 Literature Choset, Lynch,

More information

Impact of Workspace Decompositions on Discrete Search Leading Continuous Exploration (DSLX) Motion Planning

Impact of Workspace Decompositions on Discrete Search Leading Continuous Exploration (DSLX) Motion Planning 2008 IEEE International Conference on Robotics and Automation Pasadena, CA, USA, May 19-23, 2008 Impact of Workspace Decompositions on Discrete Search Leading Continuous Exploration (DSLX) Motion Planning

More information

Protein Flexibility Predictions

Protein Flexibility Predictions Protein Flexibility Predictions Using Graph Theory by D.J. Jacobs, A.J. Rader et. al. from the Michigan State University Talk by Jan Christoph, 19th January 2007 Outline 1. Introduction / Motivation 2.

More information

Specialized PRM Trajectory Planning For Hyper-Redundant Robot Manipulators

Specialized PRM Trajectory Planning For Hyper-Redundant Robot Manipulators Specialized PRM Trajectory Planning For Hyper-Redundant Robot Manipulators MAHDI F. GHAJARI and RENE V. MAYORGA Department of Industrial Systems Engineering University of Regina 3737 Wascana Parkway, Regina,

More information

Robot Motion Planning

Robot Motion Planning Robot Motion Planning James Bruce Computer Science Department Carnegie Mellon University April 7, 2004 Agent Planning An agent is a situated entity which can choose and execute actions within in an environment.

More information

6.141: Robotics systems and science Lecture 9: Configuration Space and Motion Planning

6.141: Robotics systems and science Lecture 9: Configuration Space and Motion Planning 6.141: Robotics systems and science Lecture 9: Configuration Space and Motion Planning Lecture Notes Prepared by Daniela Rus EECS/MIT Spring 2011 Figures by Nancy Amato, Rodney Brooks, Vijay Kumar Reading:

More information

A Slicing Connection Strategy for Constructing PRMs in High-Dimensional Cspaces

A Slicing Connection Strategy for Constructing PRMs in High-Dimensional Cspaces A Slicing Connection Strategy for Constructing PRMs in High-Dimensional Cspaces Abstract - This paper presents a connection strategy for PRM-based motion planning in high-dimensional configuration spaces.

More information

Adaptive Tuning of the Sampling Domain for Dynamic-Domain RRTs

Adaptive Tuning of the Sampling Domain for Dynamic-Domain RRTs Adaptive Tuning of the Sampling Domain for Dynamic-Domain RRTs Léonard Jaillet Anna Yershova LAAS-CNRS 7, Avenue du Colonel Roche 31077 Toulouse Cedex 04,France {ljaillet, nic}@laas.fr Steven M. LaValle

More information

Dynamic-Domain RRTs: Efficient Exploration by Controlling the Sampling Domain

Dynamic-Domain RRTs: Efficient Exploration by Controlling the Sampling Domain Dynamic-Domain RRTs: Efficient Exploration by Controlling the Sampling Domain Anna Yershova Léonard Jaillet Department of Computer Science University of Illinois Urbana, IL 61801 USA {yershova, lavalle}@uiuc.edu

More information

Introduction to State-of-the-art Motion Planning Algorithms. Presented by Konstantinos Tsianos

Introduction to State-of-the-art Motion Planning Algorithms. Presented by Konstantinos Tsianos Introduction to State-of-the-art Motion Planning Algorithms Presented by Konstantinos Tsianos Robots need to move! Motion Robot motion must be continuous Geometric constraints Dynamic constraints Safety

More information

Adaptive Neighbor Connection using Node Characterization

Adaptive Neighbor Connection using Node Characterization Adaptive Neighbor Connection using Node Characterization Chinwe Ekenna, Shawna Thomas, Nancy M. Amato No Institute Given Abstract. Sampling-based motion planning has been successful in planning the motion

More information

Transition-based RRT for Path Planning in Continuous Cost Spaces

Transition-based RRT for Path Planning in Continuous Cost Spaces Transition-based RRT for Path Planning in Continuous Cost Spaces Léonard Jaillet, Juan Cortés and Thierry Siméon Abstract This paper presents a new method called Transition-based RRT (T-RRT) for path planning

More information

Path Deformation Roadmaps

Path Deformation Roadmaps Path Deformation Roadmaps Léonard Jaillet and Thierry Siméon LAAS-CNRS, Toulouse, France, {ljaillet,nic}@laas.fr Abstract: This paper describes a new approach to sampling-based motion planning with PRM

More information

Collided Path Replanning in Dynamic Environments Using RRT and Cell Decomposition Algorithms

Collided Path Replanning in Dynamic Environments Using RRT and Cell Decomposition Algorithms Collided Path Replanning in Dynamic Environments Using RRT and Cell Decomposition Algorithms Ahmad Abbadi ( ) and Vaclav Prenosil Department of Information Technologies, Faculty of Informatics, Masaryk

More information

Path Planning for a Robot Manipulator based on Probabilistic Roadmap and Reinforcement Learning

Path Planning for a Robot Manipulator based on Probabilistic Roadmap and Reinforcement Learning 674 International Journal Jung-Jun of Control, Park, Automation, Ji-Hun Kim, and and Systems, Jae-Bok vol. Song 5, no. 6, pp. 674-680, December 2007 Path Planning for a Robot Manipulator based on Probabilistic

More information

Mobile Robot Path Planning in Static Environments using Particle Swarm Optimization

Mobile Robot Path Planning in Static Environments using Particle Swarm Optimization Mobile Robot Path Planning in Static Environments using Particle Swarm Optimization M. Shahab Alam, M. Usman Rafique, and M. Umer Khan Abstract Motion planning is a key element of robotics since it empowers

More information

Guiding Sampling-Based Tree Search for Motion Planning with Dynamics via Probabilistic Roadmap Abstractions

Guiding Sampling-Based Tree Search for Motion Planning with Dynamics via Probabilistic Roadmap Abstractions Guiding Sampling-Based Tree Search for Motion Planning with Dynamics via Probabilistic Roadmap Abstractions Duong Le and Erion Plaku Abstract This paper focuses on motion-planning problems for high-dimensional

More information

Planning Motion in Completely Deformable Environments

Planning Motion in Completely Deformable Environments Planning Motion in Completely Deformable Environments Samuel Rodríguez sor8786@cs.tamu.edu Jyh-Ming Lien neilien@cs.tamu.edu Nancy M. Amato amato@cs.tamu.edu Abstract Though motion planning has been studied

More information

Combinatorial Problems on Strings with Applications to Protein Folding

Combinatorial Problems on Strings with Applications to Protein Folding Combinatorial Problems on Strings with Applications to Protein Folding Alantha Newman 1 and Matthias Ruhl 2 1 MIT Laboratory for Computer Science Cambridge, MA 02139 alantha@theory.lcs.mit.edu 2 IBM Almaden

More information

Sampling Techniques for Probabilistic Roadmap Planners

Sampling Techniques for Probabilistic Roadmap Planners Sampling Techniques for Probabilistic Roadmap Planners Roland Geraerts Mark H. Overmars Institute of Information and Computing Sciences Utrecht University, the Netherlands Email: {roland,markov}@cs.uu.nl.

More information

Path Planning using Probabilistic Cell Decomposition

Path Planning using Probabilistic Cell Decomposition Path Planning using Probabilistic Cell Decomposition Frank Lingelbach Centre for Autonomous Systems Royal Institute of Technology SE-100 44 Stockholm, Sweden Email: frank.lingelbach@s3.kth.se Abstract

More information

Probabilistic Methods for Kinodynamic Path Planning

Probabilistic Methods for Kinodynamic Path Planning 16.412/6.834J Cognitive Robotics February 7 th, 2005 Probabilistic Methods for Kinodynamic Path Planning Based on Past Student Lectures by: Paul Elliott, Aisha Walcott, Nathan Ickes and Stanislav Funiak

More information

The Bridge Test for Sampling Narrow Passages with Probabilistic Roadmap Planners

The Bridge Test for Sampling Narrow Passages with Probabilistic Roadmap Planners SUBMITTED TO IEEE International Conference on Robotics & Automation, 2003 The Bridge Test for Sampling Narrow Passages with Probabilistic Roadmap Planners David Hsu Tingting Jiang Ý John Reif Ý Zheng Sun

More information

Probabilistic Roadmap Motion Planning for Deformable Objects

Probabilistic Roadmap Motion Planning for Deformable Objects Probabilistic Roadmap Motion Planning for Deformable Objects O. Burchan Bayazit Jyh-Ming Lien Nancy M. Amato Department of Computer Science Texas A&M University {burchanb,neilien,amato}@cs.tamu.edu Abstract

More information

SUN ET AL.: NARROW PASSAGE SAMPLING FOR PROBABILISTIC ROADMAP PLANNING 1. Cover Page

SUN ET AL.: NARROW PASSAGE SAMPLING FOR PROBABILISTIC ROADMAP PLANNING 1. Cover Page SUN ET AL.: NARROW PASSAGE SAMPLING FOR PROBABILISTIC ROADMAP PLANNING 1 Cover Page Paper Type: Regular Paper Paper Title: Narrow Passage Sampling for Probabilistic Roadmap Planning Authors Names and Addresses:

More information

Sampling-Based Motion Planning

Sampling-Based Motion Planning Sampling-Based Motion Planning Pieter Abbeel UC Berkeley EECS Many images from Lavalle, Planning Algorithms Motion Planning Problem Given start state x S, goal state x G Asked for: a sequence of control

More information

Path Planning. Jacky Baltes Dept. of Computer Science University of Manitoba 11/21/10

Path Planning. Jacky Baltes Dept. of Computer Science University of Manitoba   11/21/10 Path Planning Jacky Baltes Autonomous Agents Lab Department of Computer Science University of Manitoba Email: jacky@cs.umanitoba.ca http://www.cs.umanitoba.ca/~jacky Path Planning Jacky Baltes Dept. of

More information

COMPLETE AND SCALABLE MULTI-ROBOT PLANNING IN TUNNEL ENVIRONMENTS. Mike Peasgood John McPhee Christopher Clark

COMPLETE AND SCALABLE MULTI-ROBOT PLANNING IN TUNNEL ENVIRONMENTS. Mike Peasgood John McPhee Christopher Clark COMPLETE AND SCALABLE MULTI-ROBOT PLANNING IN TUNNEL ENVIRONMENTS Mike Peasgood John McPhee Christopher Clark Lab for Intelligent and Autonomous Robotics, Department of Mechanical Engineering, University

More information

Molecular Distance Geometry and Atomic Orderings

Molecular Distance Geometry and Atomic Orderings Molecular Distance Geometry and Atomic Orderings Antonio Mucherino IRISA, University of Rennes 1 Workshop Set Computation for Control ENSTA Bretagne, Brest, France December 5 th 2013 The Distance Geometry

More information

Algorithms for Sensor-Based Robotics: Sampling-Based Motion Planning

Algorithms for Sensor-Based Robotics: Sampling-Based Motion Planning Algorithms for Sensor-Based Robotics: Sampling-Based Motion Planning Computer Science 336 http://www.cs.jhu.edu/~hager/teaching/cs336 Professor Hager http://www.cs.jhu.edu/~hager Recall Earlier Methods

More information

Roadmap-Based Level Clearing of Buildings

Roadmap-Based Level Clearing of Buildings Roadmap-Based Level Clearing of Buildings Samuel Rodriguez and Nancy M. Amato Parasol Lab, Dept. Computer Science and Engineering, Texas A&M University sor8786@neo.tamu.edu,amato@tamu.edu Abstract. In

More information

ECE276B: Planning & Learning in Robotics Lecture 5: Configuration Space

ECE276B: Planning & Learning in Robotics Lecture 5: Configuration Space ECE276B: Planning & Learning in Robotics Lecture 5: Configuration Space Lecturer: Nikolay Atanasov: natanasov@ucsd.edu Teaching Assistants: Tianyu Wang: tiw161@eng.ucsd.edu Yongxi Lu: yol070@eng.ucsd.edu

More information

Principles of Robot Motion

Principles of Robot Motion Principles of Robot Motion Theory, Algorithms, and Implementation Howie Choset, Kevin Lynch, Seth Hutchinson, George Kantor, Wolfram Burgard, Lydia Kavraki, and Sebastian Thrun A Bradford Book The MIT

More information

Planning. Daniel Vallejo Christopher Jones Nancy M. Amato. Texas A&M University. Abstract

Planning. Daniel Vallejo Christopher Jones Nancy M. Amato. Texas A&M University. Abstract An Adaptive Framework for `Single Shot' Motion Planning Daniel Vallejo Christopher Jones Nancy M. Amato dvallejo@cs.tamu.edu cvj3341@cs.tamu.edu amato@cs.tamu.edu Technical Report 99-024 Department of

More information

INCREASING THE CONNECTIVITY OF PROBABILISTIC ROADMAPS VIA GENETIC POST-PROCESSING. Giuseppe Oriolo Stefano Panzieri Andrea Turli

INCREASING THE CONNECTIVITY OF PROBABILISTIC ROADMAPS VIA GENETIC POST-PROCESSING. Giuseppe Oriolo Stefano Panzieri Andrea Turli INCREASING THE CONNECTIVITY OF PROBABILISTIC ROADMAPS VIA GENETIC POST-PROCESSING Giuseppe Oriolo Stefano Panzieri Andrea Turli Dip. di Informatica e Sistemistica, Università di Roma La Sapienza, Via Eudossiana

More information

Adaptive Neighbor Connection for PRMs: A Natural Fit for Heterogeneous Environments and Parallelism

Adaptive Neighbor Connection for PRMs: A Natural Fit for Heterogeneous Environments and Parallelism Adaptive Neighbor Connection for PRMs: A Natural Fit for Heterogeneous Environments and Parallelism Chinwe Ekenna, Sam Ade Jacobs, Shawna Thomas, Nancy M. Amato Abstract Probabilistic Roadmap Methods (PRMs)

More information

Inverse Kinematics (IK)

Inverse Kinematics (IK) Inverse Kinematics Inverse Kinematics (IK) Given a kinematic chain (serial linkage), the position/orientation of one end relative to the other q 4 q (closed chain), find the values 5 of the joint parameters

More information

On the Probabilistic Foundations of Probabilistic Roadmap Planning

On the Probabilistic Foundations of Probabilistic Roadmap Planning On the Probabilistic Foundations of Probabilistic Roadmap Planning David Hsu * Jean-Claude Latombe + Hanna Kurniawati * * Department of Computer Science, National University of Singapore, Singapore + Department

More information

Motion planning is a branch of computer science concentrating upon the computation of

Motion planning is a branch of computer science concentrating upon the computation of Motion Planning for Skateboard-like Robots in Dynamic Environments by Salik Syed Introduction Motion planning is a branch of computer science concentrating upon the computation of paths for robots or digital

More information

Visualization of pinfold simulations

Visualization of pinfold simulations Visualization of pinfold simulations Sebastian Pötzsch Faculty of Mathematics and Computer Science University of Leipzig Overview 1. Introduction 1.1 Protein folding problem 1.2 HP-model 1.3 Pinfold simulation

More information

Kinodynamic Motion Planning with Hardware Demonstrations

Kinodynamic Motion Planning with Hardware Demonstrations Kinodynamic Motion Planning with Hardware Demonstrations Ioan A. Şucan Jonathan F. Kruse Mark Yim Lydia E. Kavraki Abstract This paper provides proof-of-concept that stateof-the-art sampling-based motion

More information

Nonholonomic motion planning for car-like robots

Nonholonomic motion planning for car-like robots Nonholonomic motion planning for car-like robots A. Sánchez L. 2, J. Abraham Arenas B. 1, and René Zapata. 2 1 Computer Science Dept., BUAP Puebla, Pue., México {aarenas}@cs.buap.mx 2 LIRMM, UMR5506 CNRS,

More information