arXiv subjects

Hamza Anwar

Publications and source records attributed to Hamza Anwar.

4 recordsLinked to original sources

arXiv2021

Comprehensive Energy Footprint Benchmarking Algorithm for Electrified Powertrains

Autonomy and electrification in automotive control systems have made modern-day powertrains one of the most complex cyber-physical systems. This paper presents a benchmark algorithm to quantify the performance of complex automotive systems exhibiting mechanical, electrical, and thermal interactions at various time-scales. Traditionally Dynamic Programming has been used for benchmarking the performance, however, it fails to deliver results for system with higher number of states and control lever due to curse of dimensionality. We propose "PS3", a three-step algorithm for mixed-integer nonlinear optimal control problems with application to powertrain energy management. PS3 uses pseudo-spectral collocation theory for highly accurate modeling of dynamics. Based on the validated powertrain component models, we have addressed simultaneous optimization of electrical (SOC), vehicular (eco-driving) and thermal (after-treatment and battery temperatures) dynamics along with an integer (gear and engine on/off) control and its corresponding (dwell-time) constraints. PS3 is used to solve such large-scale powertrain problems having fast and slow dynamic states, discontinuous behaviors, non-differentiable and linearly interpolated 1-D and 2-D maps, as well as combinatorial constraints. Five case study powertrain control problems are given to benchmark the accuracy and computational effort against Dynamic Programming. Our analysis shows that this algorithm does not scale computational burden as Dynamic Programming does, and can handle highly complex interactions that occur in modern-day powertrains, without compromising nonlinear and complex plant modeling.

eess.SY↗

arXiv2021

Comprehensive Energy Footprint Benchmarking of Strong Parallel Electrified Powertrain

This work presents comprehensive energy management and in-depth energy footprint analysis of an electrified strong parallel commercial vehicle. We use the PS3 framework, validated real-world powertrain system models, and Pareto-optimal analysis to optimize fuel consumption and harmful pollutant emissions. The approach involves dynamic optimization of 13 states and 4 control levers with complex interactions between multiple subsystems for a parallel hybrid electric pick-up and delivery truck. These subsystems exhibit thermal, electrical, and mechanical dynamics at different time scales, and contain kinematic and combinatorial constraints, integer- and real-valued variables, interpolated look-up tables, and data maps. A Pareto-optimal solution is found by carefully optimizing fuel and NOx emissions to understand the energy footprint of the electrified powertrain. The presented results exhibit rich analysis and complex interactions among the powertrain subsystems to unearth a 7% improvement in its fuel consumption and 29% pollutant NOx reduction when compared to solution from a coarsely modeled powertrain system.

eess.SY↗

arXiv2018

ADMM-based Networked Stochastic Variational Inference

Owing to the recent advances in "Big Data" modeling and prediction tasks, variational Bayesian estimation has gained popularity due to their ability to provide exact solutions to approximate posteriors. One key technique for approximate inference is stochastic variational inference (SVI). SVI poses variational inference as a stochastic optimization problem and solves it iteratively using noisy gradient estimates. It aims to handle massive data for predictive and classification tasks by applying complex Bayesian models that have observed as well as latent variables. This paper aims to decentralize it allowing parallel computation, secure learning and robustness benefits. We use Alternating Direction Method of Multipliers in a top-down setting to develop a distributed SVI algorithm such that independent learners running inference algorithms only require sharing the estimated model parameters instead of their private datasets. Our work extends the distributed SVI-ADMM algorithm that we first propose, to an ADMM-based networked SVI algorithm in which not only are the learners working distributively but they share information according to rules of a graph by which they form a network. This kind of work lies under the umbrella of `deep learning over networks' and we verify our algorithm for a topic-modeling problem for corpus of Wikipedia articles. We illustrate the results on latent Dirichlet allocation (LDA) topic model in large document classification, compare performance with the centralized algorithm, and use numerical experiments to corroborate the analytical results.

cs.LG↗

arXiv2017

Minimax Game-Theoretic Approach to Multiscale H-infinity Optimal Filtering

Sensing in complex systems requires large-scale information exchange and on-the-go communications over heterogeneous networks and integrated processing platforms. Many networked cyber-physical systems exhibit hierarchical infrastructures of information flows, which naturally leads to a multi-level tree-like information structure in which each level corresponds to a particular scale of representation. This work focuses on the multiscale fusion of data collected at multiple levels of the system. We propose a multiscale state-space model to represent multi-resolution data over the hierarchical information system and formulate a multi-stage dynamic zero-sum game to design a multi-scale $H_{\infty}$ robust filter. We present numerical experiments for one and two-dimensional signals and provide a comparative analysis of the minimax filter with the standard Kalman filter to show the improvement in signal-to-noise ratio (SNR).

eess.SY↗