Clark School Home UMD

June 2017

Refinement Planning and Acting

Dana Nau, Malik Ghallab and Pablo Traverso

Image: A simple example of an actor, its capabilities, and its environment.

“Refinement Planning and Acting” advances understanding of the problems that emerge when integrating deliberative acting with automated planning. A key challenge in integrating acting with planning is the discrepancy between the operational models needed for acting and the descriptive models needed for planning. This paper gives an overview of Sequential Refinement Planning Engine (SeRPE), a planning algorithm designed to be integrated with Refining Acting Engine (RAE). Although RAE’s refinement methods were designed as operational models, they still can be used for planning. The authors’ new planning algorithm, SeRPE, was written to do that. SeRPE still needs to use descriptive action models in place of RAE’s commands to the execution platform—but since it can use RAE’s refinement methods, this makes it much easier to maintain consistency between RAE and SeRPE.

Fifth Annual Conference on Advances in Cognitive Systems, May 2017. Invited paper.  A PDF of the paper can be downloaded here.


An Inexact Sample Average Approximation Approach for the Stochastic Connected Facility Location Problem

M. Gisela Bardossy and S. Raghavan

Image: Stochastic connected facility location problem example

In “An Inexact Sample Average Approximation Approach for the Stochastic Connected Facility Location Problem” the authors broaden the scope of the sample average approximation (SAA) approach, a widely used technique based on Monte-Carlo simulation, that is often applied to large-scale stochastic optimization problems. They show how even without solving the sample problems to optimality, by combining a heuristic and a lower bounding approach, high-quality solutions with tight confidence bounds on the optimal solution value can be obtained. Their computational results demonstrate the effectiveness of this “inexact SAA approach.”

NETWORKS, Vol. 000(00), 000–000 2017. April 28, 2017.  The article can be viewed and downloaded here.



A platform for in situ Raman and stress characterizations of V2O5 cathode using MEMS device

Hyun Jung, Konstantinos Gerasoloulos, A. Alec Talin, Reza Ghodssi


Image: (a) Schematic of the coin cell parts (from top to bottom: cathode cap, adhesive tape, MEMS optical stress sensor, separator, lithium anode, stainless steel, spring, and anode cap) for in situ experiment and simplified schematic diagrams of the experimental setup showing the location of the coin cell under test relative to the Raman microscope. (b) Optical image of the experimentally obtained interference pattern from the membrane. (c) Experimentally obtained Raman spectrum of as-deposited V2O5 thin film electrode underneath the glass wafer and membrane.

The authors present an in situ approach for the characterization of lithium intercalation/deintercalation in thin-film Li-ion battery electrodes. The method allows simultaneous measurement of microstructural changes during lithiation using micro-Raman (μRaman) spectroscopy in parallel with stress changes via optical interferometry. They observe evolution in the microstructure and stress in the various crystal phases in the Li-V-O system, including both reversible and irreversible phase transitions. They also correlate spectral shifts in certain Raman active modes with changes in the electrode stress, and thus confirm previously hypothesized origins of these observations. Ultimately, the combined stress/μRaman in situ technique can be leveraged as a characterization platform for a wide variety of electrode materials for advancing battery performance.

Electrochimica Acta Volume 242, 10 July 2017, Pages 227–239  View or download the article here.


Geometric decompositions of collective motion

Matteo Mischiati, P. S. Krishnaprasad

Image: Ensemble inertia tensor of a collective (point cloud) visualized as an ellipsoid located at the center of mass (Figure 1 of paper – copyright Royal Society)

The authors devise a new fiber bundle structure to describe the changing forms of flocks and underlying kinematic modes. The journal website includes supplementary material that provides details on how the concept of ensemble inertia tensor described by just 6 parameters, is the key to the new structure. Under certain generic conditions, flock data can be factored into such a tensor and coordinates on a Stiefel manifold, a space that has played an important role in various areas of mathematics and applied science. The modes are arrived at in a principled manner via modern tools from the field of differential geometry. Using notions of energy attributed to various types of collective movement, the authors shed new light on data obtained by previous researchers from small flocks of pigeons in free flight over large distances. You can read an ISR story about the research here.

Proceedings of the Royal Society A, Published 26 April 2017.DOI: 10.1098/rspa.2016.0571  A PDF of the article can be downloaded here.


Group Cooperation with Optimal Resource Allocation in Wireless Powered Communication Networks

Ke Xiong, Chen Chen, Gang Qu, Pingyi Fan, Khaled Ben Letaief

Problem scenario: Communication Group 1 has sufficient bandwidth but limited transmission power, Group 2 has sufficient power but limited bandwidth. Our proposed 4-phase cooperative transmission protocol to solve the problem.

This paper considers a wireless powered communication network (WPCN) with group cooperation, where two communication groups cooperate with each other via wireless power transfer and time sharing to fulfill their expected information delivering and achieve “win-win” collaboration. To explore the system performance limits, the authors formulate optimization problems to maximize the weighted sum-rat e (WSR) and minimize the total consumed power. The time assignment, beamforming vector and power allocation are jointly optimized under available power and quality of service requirement constraints of both the groups. For the WSR-maximization, both fixed and flexible power scenarios are investigated. As all problems are non-convex and have no known solution methods, the authors solve them by using proper variable substitutions and the semi-definite relaxation. They theoretically prove that their proposed solution method guarantees the global optimum for each problem. Numerical results are presented to show the system performance behaviors, which provide some useful insights for future WPCN design. It shows that in such a group cooperation-aware WPCN, optimal time assignment has the greatest effect on the system performance than other factors.

IEEE Transactions on Wireless Communications Volume 16, Issue 6, June 2017 The paper can be viewed or downloaded here.


Adaptive System Optimization using Random Directions Stochastic Approximation

L.A. Prashanth, Shalabh Bhatnagar, Michael Fu and Steve Marcus


Image: Road network used in the experiments

The authors present new algorithms for simulation optimization using random directions stochastic approximation (RDSA). These include first-order (gradient) as well as second-order (Newton) schemes. They incorporate both continuous-valued as well as discrete-valued perturbations into both types of algorithms. The former are chosen to be independent and identically distributed (i.i.d.) symmetric uniformly distributed random variables (r.v.), while the latter are i.i.d. asymmetric Bernoulli r.v.s. Their Newton algorithm, with a novel Hessian estimation scheme, requires N-dimensional perturbations and three loss measurements per iteration, whereas the simultaneous perturbation Newton search algorithm of [1] requires 2N-dimensional perturbations and four loss measurements per iteration. They prove the asymptotic unbiasedness of both gradient and Hessian estimates and asymptotic (strong) convergence for both first-order and second-order schemes. We also provide asymptotic normality results, which in particular establish that the asymmetric Bernoulli variant of Newton RDSA method is better than 2SPSA of [1]. Numerical experiments are used to validate the theoretical results.

IEEE Transactions on Automatic Control, Vol. 62, No. 5, May 2017.  A PDF of the article can be downloaded here.



Image: At left, the nanosponge is loaded with particles that are deposited by sliding the sponge along the surface. At right, an empty nanosponge “eraser” that can wet and erase dried up materials. In the center, the real image of the sponge magnified 4000 times with an electron microscope. CREDIT: Nanoscience Exploration Research and Development (NERD) Lab, SIUC.

As a grad student, Pradeep Rajasekaran (currently a postdoc with Reza Ghodssi) was part of a team that developed a tiny “nanosponge” editing tool that could make a big impact in semiconductor, electronics and even biotechnology manufacturing—industries that use lithographic printing techniques in manufacturing. The work has just been published in Science Advances as “Polymeric Lithography Editor: Editing Lithographic Errors with Nanoporous Polymeric Probes.” Read the ISR story here and view or download the Sciences Advances paper here.