Mentar Mahmudi

Designing antibiotic cycling strategies by determining and understanding local adaptive landscapes.

The evolution of antibiotic resistance among bacteria threatens our continued ability to treat infectious diseases. The need for sustainable strategies to cure bacterial infections has never been greater. So far, all attempts to restore susceptibility after resistance has arisen have been unsuccessful, including restrictions on prescribing [1] and antibiotic cycling [2], [3]. Part of the problem may be that those efforts have implemented different classes of unrelated antibiotics, and relied on removal of resistance by random loss of resistance genes from bacterial populations (drift). Here, we show that alternating structurally similar antibiotics can restore susceptibility to antibiotics after resistance has evolved. We found that the resistance phenotypes conferred by variant alleles of the resistance gene encoding the TEM β-lactamase (bla TEM) varied greatly among 15 different β-lactam antibiotics. We captured those differences by characterizing complete adaptive landscapes for the resistance alleles bla TEM-50 and bla TEM-85, each of which differs from its ancestor bla TEM-1 by four mutations. We identified pathways through those landscapes where selection for increased resistance moved in a repeating cycle among a limited set of alleles as antibiotics were alternated. Our results showed that susceptibility to antibiotics can be sustainably renewed by cycling structurally similar antibiotics. We anticipate that these results may provide a conceptual framework for managing antibiotic resistance. This approach may also guide sustainable cycling of the drugs used to treat malaria and HIV.

Planning humanlike actions in blending spaces

We introduce an approach for enabling sampling-based planners to compute motions with humanlike appearance. The proposed method is based on a space of blendable example motions collected by motion capture. This space is explored by a sampling-based planner that is able to produce motions around obstacles while keeping solutions similar to the original examples. The results therefore largely maintain the humanlike characteristics observed in the example motions. The method is applied to generic upper-body actions and is complemented by a locomotion planner that searches for suitable body placements for executing upper-body actions successfully. As a result, our overall multi-modal planning method is able to automatically coordinate whole-body motions for action execution among obstacles, and the produced motions remain similar to example motions given as input to the system.

Analyzing Locomotion Synthesis with Feature-Based Motion Graphs

We propose feature-based motion graphs for realistic locomotion synthesis among obstacles. Among several advantages, feature-based motion graphs achieve improved results in search queries, eliminate the need of postprocessing for foot skating removal, and reduce the computational requirements in comparison to traditional motion graphs. Our contributions are threefold. First, we show that choosing transitions based on relevant features significantly reduces graph construction time and leads to improved search performances. Second, we employ a fast channel search method that confines the motion graph search to a free channel with guaranteed clearance among obstacles, achieving faster and improved results that avoid expensive collision checking. Lastly, we present a motion deformation model based on Inverse Kinematics applied over the transitions of a solution branch. Each transition is assigned a continuous deformation range that does not exceed the original transition cost threshold specified by the user for the graph construction. The obtained deformation improves the reachability of the feature-based motion graph and in turn also reduces the time spent during search. The results obtained by the proposed methods are evaluated and quantified, and they demonstrate significant improvements in comparison to traditional motion graph techniques.

Precomputed motion maps for unstructured motion capture

We present in this paper a solution for extracting high-quality motions from unstructured motion capture databases at interactive rates. The proposed solution is based on automatically-built motion graphs, and offers two key contributions. First, we show how precomputed expansion trees (or motion maps) coupled with new heuristics and backtracking techniques are able to significantly improve the time taken to search for motions satisfying user constraints. Second, we show that when feature-based transitions are employed for constructing the underlying motion graph, the connectivity of motion maps is greatly increased, allowing the overall method to perform search and synthesis at interactive frame rates. We demonstrate the effectiveness of our approach with the problem of extracting path-following motions around obstacles from a motion graph structure at interactive performances.

Feature-Based locomotion with inverse branch kinematics

We propose a novel Inverse Kinematics based deformation method that introduces flexibility and parameterization to motion graphs without degrading the quality of the synthesized motions. Our method deforms the transitions of a motion graph-like structure by first assigning to each transition a continuous rotational range that guarantees not to exceed the predefined global transition cost threshold. The deformation procedure improves the reachability of motion graphs to precise locations and consequently reduces the time spent during search. Furthermore, our method includes a new motion graph construction method based on geometrical segmentation features, and employs a fast triangulation based search pruning technique that confines the search to a free channel and avoids expensive collision checking. The results obtained by the proposed methods were evaluated and quantified, and they demonstrate significant improvements in comparison with traditional motion graph approaches.

Multi-modal data-driven motion planning and synthesis

We present a new approach for whole-body motion synthesis that is able to generate high-quality motions for challenging mobile-manipulation scenarios. Our approach decouples the problem in specialized locomotion and manipulation skills, and proposes a multi-modal planning scheme that explores the search space of each skill together with the possible transition points between skills. In order to achieve high-quality results the locomotion skill is designed to be fully data-driven, while manipulation skills can be algorithmic or data-driven according to data availability and the complexity of the environment. Our method is able to automatically generate complex motions with precise manipulation targets among obstacles and in coordination with locomotion.

Fast path planning using motion graphs

We present our ongoing work on a novel method for navigating humanoid characters through an unstructured complex environment by the help of a fast path planning algorithm integrated with motion graphs. Our main goal is to achieve characters showing human-like motions provided by motion capture and at the same time being able to compose complex realistic motions for executing path following in real-time. Current methods that solve the path planning problem with motion graphs are not real-time and do not guarantee that the humanoid character is walking on a collision-free path. Furthermore there are no guarantees on how close the composed motion is to the path being followed. Our method addresses these issues by employing a fast path planning and execution mechanism whenever possible and only employing motion graph search when needed. Our method is directly applicable to video games.

Artist-oriented 3D character posing from 2D strokes

In this paper we present an intuitive tool suitable for 2D artists using touch-enabled pen tablets. An artist-oriented tool should be easy-to-use, real-time, versatile, and locally refinable. Our approach uses an interactive system for 3D character posing from 2D strokes. We employ a closed-form solution for the 2D strokes to 3D skeleton registration problem. We first construct an intermediate 2D stroke representation by extracting local features using meaningful heuristics. Then, we match 2D stroke segments to 3D bones. Finally, 3D bones are carefully realigned with the matched 2D stroke segments while enforcing important constraints such as bone rigidity and depth. Our technique is real-time and has a linear time complexity. It is versatile, as it works with any type of 2D stroke and 3D skeleton input. Finally, thanks to its coarse-to-fine design, it allows users to perform local refinements and thus keep full control over the final results. We demonstrate that our system is suitable for 2D artists using touch-enabled pen tablets by posing 3D characters with heterogeneous topologies (bipeds, quadrupeds, hands) in real-time. Graphical abstractDisplay Omitted HighlightsFast artist-oriented 2D stroke driven incremental refinement of 3D poses.Novel linear-time stroke-chain matching and feature extraction techniques.Modular joint rotation computation preserving joint depth and bone rigidity.Intuitive integration stroke-driven 3D pose design with Inverse Kinematics.Usability study with 22 subjects resulting in high SUS score.