Kevin Angers

I’m a fourth-year undergraduate student at the University of Toronto in Engineering Science, majoring in Robotics with a minor in Machine Intelligence.

I am currently working toward my undergraduate thesis at the Robot Vision and Learning (RVL) Lab supervised by Prof. Florian Shkurti (affiliations: UofT Robotics Institute, Vector Institute, Acceleration Consortium). I am working on generalizable end-to-end robot manipulation focusing on task-oriented, multi-stage problems.

I have previously done research at the Matter Lab supervised by Prof. Alán Aspuru-Guzik (affiliations: Vector Institute, Acceleration Consortium, CIFAR), Prof. Milica Radisic (affiliations: UofT BME, UHN, Acceleration Consortium), and Prof. Florian Shkurti on general-purpose robotics for laboratory automation, and our publication “RoboCulture” is under review at Matter.

I am also a member of aUToronto, the University of Toronto’s self-driving car team, as a member of the 3D-object detection subteam.

I am a team capitan of the University of Toronto Varsity Blues Baseball Team (3x OUA Champions) and two-time OUA Most Valuable Pitcher.

Education

University of Toronto

BASc. in Engineering Science | 2020 - 2025
Robotics Major, Machine Intelligence Minor

Publications

RoboCulture: A Robotic Platform for Automated Biological Experimentation
Kevin Angers, Kourosh Darvish, Naruki Yoshikawa, Sargol Okhovatian, Dawn Bannerman, Ilya Yakavets, Florian Shkurti, Milica Radisic, Alán Aspuru-Guzik
Matter, 2024 (Submitted)

PAPER / PROJECT PAGE (Coming Soon!)/ CODE (Coming Soon!) / POSTER

RoboCulture is a robotics platform enabling the automation of liquid handling-based biological experiments using a Franka Panda robot manipulator. Computer vision strategies are used to facilitate robust pipetting amidst small targets at unknown positions.

Projects

PitChart - iOS App for Baseball Pitch Charting

APP STORE LINK

iOS app using SwiftUI for tracking sequences of pitches for player trend analysis with the UofT Baseball team. Record pitch location, type, result and type of contact. Pitch sequences are organized into at bats; track the pitchers and the batters throughout a game. Add new games and track players' trends and matchups throughout a season.

Motion Planning on KUKA Manipulator

Course Lab Project – ECE470 (Robot Modeling and Control). Code avaliable upon request

Implemented the artificial potential algorithm in MATLAB, simulating attractive and repulsive forces for a KUKA manipulator allowing it to grasp and place objects at desired points while avoiding obstacles.

Cart Pendulum Control

Course Lab Project – ECE557 (Linear Control Theory). Code avaliable upon request

Designed an output feedback controller to enable a cart-pendulum system to track a square wave signal while keeping the pendulum balanced in the vertical upright configuration. Utilized state-space control design principles to design an observer for the linearized system, simulated in Simulink, and tested on a physical system using an Arduino.

Transformer Language Model from Scratch

Course Project – CSC401 (Natural Language Computing). Code avaliable upon request

Implemented a Transformer based language model from scratch in PyTorch for the task of machine translation, and trained on a corpus of English-French text for performance evaluation using the BLEU metric.

Bayesian Localization with Mobile Robot

Course Lab Project – ROB301 (Introduction to Robotics). Code avaliable upon request

Implemented Bayesian localization and Kalman filtering algorithms for accurate state predictions with noisy measurements, enabling robust navigation of a mobile robot across a topological map, simulating mail delivery.

Globe Bot – A Voice Controlled Globe

Course Project - MIE438 (Microcontrollers and Embedded Microprocessors) CODE

Ask the Globe Bot to point to any location, and it will spin and highlight the given location on the globe under a magnifying glass.

Aerial Image Recognition with CNN

Course Project – APS360 (Introduction to Machine Learning). Code avaliable upon request

Trained a convolutional neural network with PyTorch to classify aerial (satellite) imagery across 40 classes. Used transfer learning to leverage AlexNet architecture for feature extraction.