AI-Driven Upgrade for Robotic Vacuum Cleaners
As a computer science student, you often wonder when you’ll get to apply what you’ve learned in the real world. I got that chance when I worked on a project for a top engineering and technology company. The task was to develop an AI-powered depth estimator to improve robotic vacuum cleaners. This project was crucial in building my skills, understanding industry practices, and learning the importance of teamwork in software development.
Project Overview
The project aimed to enhance the company’s robotic vacuum cleaners by replacing expensive depth sensors, like LIDAR and Stereo Cameras, with a more affordable monocular camera system. Our task was to develop an AI-driven Monocular Visual-Inertial Depth Estimator for accurate depth sensing. However, since we were taking over over the project from a previous team, we spent the first phase revisiting their work and grasping the project’s fundamentals.
Understanding Real-World Processes
The main difference between academic and real-world projects is managing expectations and balancing priorities. In class, the focus is on technical skills, but in this project, we had to manage time, communicate effectively, and adapt to unexpected challenges. We began by analyzing the previous team’s work, breaking the project into smaller tasks, and setting realistic timelines. Unlike class assignments, real-world projects require detailed planning, prioritizing, and anticipating problems.
Collaborating with Stakeholders
One of the most rewarding aspects of the project was working with industry professionals. As the project lead, I had regular meetings with stakeholders, including engineers and managers. These interactions helped us align our work with the company’s larger goals. Communication was key, as we frequently updated stakeholders on challenges and sought their feedback and help to align with the project’s objectives.
Execution
Given the project’s complexity, we followed the Agile methodology to stay flexible. Our initial focus was on designing a pipeline to estimate metric-depth in real-time using a monocular camera and IMU. Optimizing the system for real-time performance on embedded hardware was especially tough, but our persistence paid off, reducing hardware costs while maintaining depth accuracy. We rigorously tested the system in various indoor environments. Each test revealed areas to improve, and we continuously refined the system to meet standards.
Lessons Learned
This project taught me much more than just technical skills. Leading a team and working closely with stakeholders gave me valuable insights into problem-solving in the real world. I learned how to lead under pressure, communicate effectively, and navigate teamwork challenges.
On the technical side, I enhanced my skills in AI, robotics, and project management, but the biggest takeaway was balancing technical knowledge with project leadership and collaboration.
Final Reflections
This project taught me the importance of leaving projects well-documented for future teams and how to handle complex situations. This project not only expanded my technical abilities but also prepared me for future software development roles, where managing challenges and working in teams are essential skills.
