Projects

Research Projects

Deep Learning for Robot Control | MASc/PhD Thesis
Advisors: Prof. Angela P. Schoellig and Dr. Mohamed K. Helwa
Institute for Aerospace Studies, University of Toronto

a) Derived platform-independent guidelines from control theory for efficiently training neural networks as system inverse dynamic models to enhance robot impromptu trajectory tracking performance; demonstrated effectiveness with experiments on quadrotor vehicles (CDC 2017, CoRL 2017)

b) Extended the inverse-learning-based approach to the challenging case of non-minimum phase systems where system inverse dynamics are unstable; experimentally verified with an inverted pendulum on a cart system (RA-L)

>> more details


Admissibility of Distributed Control Law | BASc Thesis & Summer Research
Advisor: Prof. Angela P. Schoellig
Co-advisor: Prof. Bruce A. Francis (for Summer Research)
Institute for Aerospace Studies, University of Toronto

a) Analyzed the feasibility of applying distributed control laws designed based on kinematic points to nonlinear nonholonomic robot models; experimentally verified a proposed linearization-based approach for transforming leader-follower distributed control law based on kinematic points to unicycle models

b) Derived upper bounds of tracking error for practical control designs and addressed the issue of actuation constraints in implementations


Computational Fluid Dynamics | Summer Research
Advisor: Prof. David W. Zingg
Institute for Aerospace Studies, University of Toronto

a) Researched the integrated parameterization and mesh movement algorithm for generating computational
meshes used in flow analyses and aerodynamic shape optimizations; assessed the flexibility of a B-spline-based
mesh movement algorithm for generating 3D computational meshes from 2-patch flat plates

b) Analyzed the trade-off between accuracy and computational cost of using higher-order methods for solving
the Euler’s equations in 2D inviscid flow problems


Course Projects

Neural State Estimation
Course: Computational Neuroscience
Department of Computer Science, University of Toronto

Description: Studied biologically plausible neural implementations of Bayesian inference involved in estimating time-varying quantities; analyzed characteristics of the neural state estimators derived based on the probabilistic population code and the sampling-based code to assess their plausibilities (CCN 2017)


Nonlinear State Estimation for Robotics
Course: State Estimation for Aerospace Vehicles
Institute for Aerospace Studies, University of Toronto

Description: Implemented and explored the properties of Particle Filter, Sigma-Point Kalman Filter, and Extended Kalman Filter for estimating the states of a ground robot modeled by the nonlinear unicycle model; assessed the robustness and accuracy of the estimators in challenging cases where the vision range became increasingly limited and where the range measurements were corrupted


Canada’s Next Generation Robotics
Course: Space Systems Design
Institute for Aerospace Studies, University of Toronto

Description: Collaborated with four team members to develop the design of a space manipulator with capabilities of performing autonomous station reconfiguration, capture and berthing, and maintenance tasks


Dowel Packing Machine Design
Course: Engineering Design
Institute for Aerospace Studies, University of Toronto

Description: Collaborated with three team members to design and build a portable machine that autonomously sorts and packs colour-coded wooden dowels to provided compartment containers