Project CG-63: Gantry Scanner Control System for a Mobile Scanning Rover
This team will not have a booth at the Design & Innovation Day due to confidentiality agreements.
Project Client: FPInnovations
Project Description: This project is a proof of concept for FPInnovations to demonstrate how the in-field log grading process can be automated. In the current state, logs that have been cut down are placed in a log yard, usually in cold, muddy, outdoor conditions, and workers are required to manually measure and grade each log to be sorted for the mill. This Log Gantry Scanner project aims to mitigate this by having an automated system scan each log to retrieve key log measurements such as the length and the small end and large end diameters. It is also a fully remote controlled system that can be operated by a single person.
Project HA-76: Modena Smart Home Next-Gen User Interface
Project Client: iaconicDesign Inc.
Problems: The market for smart home products is continuing to grow. Our client, iaconicDesign, wants to further its development of innovative products in the smart home market under the “Modena Smart Home” brand name. Our client wants to create multiple generations of modular smart home products to complement their current smart home light switches. For the newest generation of light switches, they want to:
– have a wall-mounted touchscreen as a controller
– have a clear display with a polished user interface
– keep retail prices well below current prices (current smart switches cost from $400-$600)
Solutions: Our team did extensive research on touch displays on the market to minimize raw material cost. The project addresses the end user’s needs by providing them with a solution to
modify their existing light system. These materials include:
– Esp32 as the controller
– 3.5 inch Screen with capacitive touch
– LCD Display
Major design contribution: Our major design contribution is composed of 2 parts; screen selection and UI design. Our team used a WDM (Weighted Decision Matrix) to decide which touch screens to buy and test for our final prototype. Criteria and weight are based on the needs of our stakeholders. Some criteria include cost, size, resolution, and shipping time. To make our UI simple and user friendly, we made the controls relatively similar to how one navigates through screens on their smartphones, with minimal taps to access each function.
Project JY-77: Online Price Monitoring Application
Project Client: Cymax Group Technologies
Project Description: On large marketplaces like Amazon, identical products are sold by multiple resellers that compete on price. In order to promote fair competition, maintain brand identity, and protect profit margins, vendor brands set a minimum price that resellers are legally able to advertise products for — known as the Minimum Advertised Price (MAP). However, with such intense competition, resellers are often motivated to sell under MAP in order to increase their sales. With such a large number of E-commerce websites, monitoring MAP violations manually is not feasible. We partnered with Cymax Group, a third-party reseller of home furnishings working with vendors to sell products on online marketplaces, and built the Online Price Monitoring Application, which solves this problem by monitoring MAP violations in an automated way. The application provides a web crawler that periodically crawls E-commerce websites for MAP violations and persists the product prices in a database. Vendors and manufacturers are able to log into a web frontend to view price metrics and any MAP violations for their products. The application also notifies them via email whenever violations occur, so they can take action as they see fit.
Project PB-19: Automating Product Ranking and Competitive Product Analysis to Achieve Success At Scale
Project Client: Esparkify
Project Description: Businesses on online marketplaces such as Amazon can generate large amounts of sales, but only if their products are ranked high and appear at the top of the marketplace’s search engine. Brands with advertising strategies not optimized for Amazon’s search engine often struggle to connect with potential customers. Specifically, it is hard to predict what keywords to use so that their products get more visibility in customers’ search results on Amazon. Our project aims to help sellers conduct market research and provide sellers with a keyword analysis tool. Given any search term, this tool recommends sellers with the highest scoring keywords based on Esparkify’s Keyword Ranking system. Automating keyword extraction and data scraping removes the time-consuming process to do a manual analysis on every product keyword. Sellers can put the recommended keywords in their product information to gain a competitive edge.
Contact Information: ubccapstone19@gmail.com
Project PB-34: Slicing-Based Debugging for Java/Android Development Environments
Project Client: UBC ECE, ReSeSS Research Lab, PI: Prof. Julia Rubin.
Project Description: Our team has worked with the ReSSeS lab to integrate an innovative debugging tool called the Slicer4J Plugin. Slicer4J is a dynamic program analysis tool launched from a command line that allows users to select a statement and take a slice of the program execution to find all the lines which impact that statement. It enables analysis of a full execution of a program with a bird’s-eye view as opposed to the moment-in-time views that are common with traditional debugging tools. The objective of this project is to produce a functional, intuitive, and simple-to-use plugin for IntelliJ IDEA that utilizes the Slicer4J tool.
Our technical challenges included:
– Interfacing plugin with the Slicer4J tool
– Processing data from Slicer4J and translating slicing result from bytecode to source code level
– Designing a user-friendly and accessible interface
– Generating an interactive Graph and Table to visualize slicing results
– Highlighting source code for relevant statements
In addition to development, we conducted a user study to evaluate the effectiveness of the Slicer4J Plugin compared with the traditional IntelliJ Debugger.
Contact information: David Fong: davidfong19@gmail.com ; Tristen Raab: tristenraab@gmail.com ; John Ramsden: john@ubc.ramsdenj.ca ; Preet Shah: pshah5000@gmail.com ; Danica Xiao: danica5471@gmail.com
Project PL-50: PropBot – Propagation data collection Robot
Project Client: UBC Radio Science Lab
Project Description: Propbot is a fully autonomous robot designed to help researchers at the UBC Radio Science Lab collect wireless propagation data more easily. Current methods to collect data involve researchers manually pushing around carts containing heavy data collection equipment, which is both a tedious and time consuming process.
The goal of Propbot is to systematize and automate the task of data collection as to make the task easier and faster for researchers. Propbot is a custom robot capable of being teleoperated by a researcher. The researcher can drive Propbot to points of interest without the physical labor involved with pushing a cart, all while moving faster. The task of data collection can be made even simpler using the integrated autonomous system, which allow researchers to upload data collection points on the custom mission command software and hand off the task of driving to the autonomous system.
This project is a continuation of a previous Capstone project, and our team focused on the implementation of a variety of key design features including autonomy system, teleoperation system, short-range obstacle avoidance, and power management.
Contact information: Amr Almoallim: amralmoallim@gmail.com ; Sajjad Al-Kazzaz: alkazzazsajjad@gmail.com ; Johanan Agarwal: johanan888@outlook.com ; Apoorv Garg: apoorvgarg17@gmail.com ; Cole Shanks: cshanks223@gmail.com
Project PN-67: End-to-End Implementation of Automatic Road Deterioration Detection for Smart City Applications
Project Client: TELUS Communications Inc.
Project Description: Have you ever hit a pothole so hard you drove with the radio off for 10 mins?
Every year Canadians spend about $3 billion to fix their cars from pothole damage. This is in addition to what governments spend on fixing roads, the city of Vancouver alone spends over $25 million annually to fix roads.
Locating and identifying road deterioration is a slow process; municipal employees or city residents need to report them manually. The purpose of this project is to help municipalities identify road deterioration in the city faster than existing processes and early on using machine learning. CloudRoad also provides a dashboard for municipal employees to view road deterioration around the city.
CloudRoad benefits Canadians as it helps reduce the unnecessary stress caused by road deterioration. Focusing on repairing the worst roads or preventing road deterioration to the point of major repairs also saves millions of dollars for Canadians
Contact Information: Daniyal Bhaila: daniyalbhaila@gmail.com (https://www.linkedin.com/in/daniyalbhaila/) ; Dryden Wiebe: dryden.wiebe@gmail.com
Project PN-68: Automatic Recommendation system for the video streaming platforms according to the users’ viewing history
Project Client: TELUS
Project Description: Telus is a telecommunications company that provides services including streaming and live television. With the aim of providing account-specific show and movie recommendations, Telus is seeking a recommendation system. Our project includes the design and development for a recommendation system prototype that Telus can use to determine the efficacy of building one in-house; as opposed to using a third-party system.
Project PN-71: End-to-End Implementation of Street Parking Detection for Smart City Applications
Project Client: TELUS
Project Description: Ever had trouble finding street parking? ParkMate is a new, state-of-the-art, camera detection system that utilizes the latest in machine learning technology to help you find open street parking ahead of time. This project was developed as part of a Smart City initiative with the goal of reducing traffic congestion and greenhouse gas emissions. ParkMate integrates directly with the car’s existing camera and navigation systems, so no additional technology is required!
Contact information: masrur.mahbub@gmail.com
Project SF-59: An ML-Based Cost Prediction Platform for Emerging Cloud Services
Project Client: UBC Vancouver / ECE Department / Cloud Infrastructure Research for Reliability, Usability, and Scalability (CIRRUS) Lab
Project Description: Serverless Price Optimization Tool (SPOT) is an open-source software-as-a-service (SaaS) that provides configuration recommendations to stable, production-level serverless applications deployed in Amazon Lambda. The resource configuration of a serverless application impacts its price and performance greatly but the optimal point is hard to find. With UBC CIRRUS Lab, our team designed and built SPOT to fill this gap in cost optimization for industry-level applications. SPOT can pull logs and configurations from a Lambda function and fit all the data to a pre-tuned model. As the function gets invoked more and generates more data, SPOT can gather all the new information, adapt the model and provide more accurate predictions.
Project SF-87: Tap to Park – Simplified Parking Permit System
Project Client: Designwerx Innovation Inc
Project Description: Currently, the pay parking systems in Vancouver are confusing, inconsistent and frustrating. Each lot seems to have its own parking metre, phone application, and regulations. Our client, Antony Hodgson at Designwerx Innovation Inc, aims to eliminate parking lot frustration with Tap To Park.
Our team worked alongside him to develop a small, streamlined device and companion phone application that makes pay parking effortless. The Tap To Park system allows you to simply park in your preferred spot, tap the button on the physical Tap To Park device located inside your vehicle, and be notified right away on its display when your session has started! This is all accomplished without pulling out your phone or having to use a physical metre. The device will connect with our companion phone application in the background and handle all the important details without you having to do a thing.
In addition, when you’re done parking, all you have to do is drive away! Our application can detect your car’s location and automatically end the session, making sure you’ve only paid for exactly the time you spent inside the lot.
We hope to eliminate the hassle of fumbling for change at the metre, trying to understand complex pay parking applications, and being charged for unused parking time, all with Tap to Park!
Project TL-70: Indoor 3D Wi-Fi heat map generation based on floor plan and mobile 2D/3D sensors
This team will not have a booth at the Design & Innovation Day due to confidentiality agreements.
Project Client: TELUS
Project Description: The objective of the project is to develop a complete platform that first allows the user to scan an indoor environment using a LiDAR sensor (of a mobile device such as iPad Pro), upload an indoor floorplan, and generate a 3D model of the space; then a Wi-Fi signal measuring API will be employed to measure the signal strength around the house. The collected signal information will be automatically overlayed (e.g., as numbers or in the form of heat map) on the 3D model.
Major design contribution: due to IP and NDA, no information will be provided.
Contact Information: Haoxiang Qi: qihaoxiang@gmail.com ; Yihan Shang: yihanshang@yahoo.com ; Yuhang Zhang: zyuhang2000@126.com ; Zihao Su: marcosu386@gmail.com ; Dhruv Patel: dhruvpatel1999@outlook.in
Project TL-74: Machine learning approaches to accelerate COVID-19 drug discovery
Project Client: Gandeeva Therapeutics Inc.
Project Description: With emerging mutations of viruses such as Omicron-COVID-19, shortening the timeline between variant identification to drug discovery is more crucial than ever before. By combining cryogenic electron microscopy (cryo-EM) and artificial intelligence we can accelerate the cryo-EM imaging pipeline, thus shortening the timeline of drug discovery.
Cryo-EM images frozen proteins at atomic resolution. The final structure is typically determined by averaging the information from <100,000 molecular images. A key step in the process is the identification of regions where these specific molecular images are located from the nearly 10 billion molecules that may be potentially present on a single specimen grid. This is achieved by imaging at very low magnifications initially, followed by progressive selection of the regions most likely to yield the best images.
While state-of-the-art methods rely on going through this process manually, we demonstrate a 10-fold acceleration of the particle imaging selection process using machine learning techniques, including convolutional neural networks and clustering. Increased efficiency and complete automation of this process yield significant dividends in speeding up the process of getting the final structure, and have immediate application to imaging of drugs complexed to proteins.
Team Members: Carmen (Hanyu) Che, Martin Chua, Eunice Kim, Paul (Yen Fu) Lin, Tony (Lingtong) Xu