The Collaborative Autonomous Robot Delivery System (CARDS) was a capstone project focused on developing a modular, self-balancing robotic delivery platform capable of operating independently or in tandem with another unit. Designed for semi-structured environments, the system integrates mechanical, electrical, and software subsystems to enable collaborative load transport in areas that traditional wheeled delivery robots struggle with. Central to the design was a bipedal robot platform equipped with a robust cycloidal gearbox, a custom leg mechanism for vertical support, and a hybrid latching system that allows two robots to physically and functionally pair together for increased payload capacity.
Built on a Jetson Orin Nano and dual Teensy 4.1 microcontroller architecture, the robot uses a dual PID control loop for self-balancing and motion stability. The software stack includes ROS2 integration, real-time motor coordination via CAN bus, and Kalman-filtered sensor fusion for precise movement. Mechanical design was optimized through 3D printing, modular aluminum framing, and embedded sensors to allow for streamlined testing, durability, and repeatability. The project culminated in a fully functioning prototype capable of balancing, collaborating via latching, and navigating complex environments, demonstrating a scalable approach to autonomous delivery.
Mechanical Design: Engineered a custom cycloidal gearbox capable of delivering >139 Nm of torque, integrated with a spring-assisted leg system for improved load-bearing and energy efficiency.
Software Integration: Developed a dual PID-based control system with sensor fusion, enabling real-time self-balancing and disturbance recovery within 2.5 seconds.
Latching Mechanism: Designed and tested a hybrid passive-active latching system using an electromagnet and gate latch, achieving over 90% successful coupling in lab conditions.
Power and Communication Systems: Implemented a 48V high-voltage power distribution and CAN bus communication architecture, optimizing power usage and enabling reliable multi-motor synchronization.
Validation & Testing: Successfully validated self-balancing performance, communication latency (<10 ms), system runtime (>30 min), and collaborative transport in both simulated and real-world testing environments.
Mechanical Design & Fabrication
Led the complete mechanical design of the robot, including the chassis, leg system, gearbox, and latching mechanism, using SolidWorks and rapid prototyping techniques.
Designed and built a custom cycloidal gearbox with a 30:1 reduction ratio, achieving 139.5 Nm of torque—exceeding project requirements.
Spearheaded the modular chassis design using aluminum extrusions and 3D-printed components, optimizing for strength, adaptability, and ease of maintenance.
Incorporated spring-assisted legs to reduce motor load and improve power efficiency and vertical mobility.
System Assembly & Integration
Personally oversaw the entire mechanical assembly of the robot, including gearbox mounting, leg installation, electrical housing, and joint connections.
Implemented heated inserts, threaded fasteners, and modular subassemblies to ensure robust and serviceable construction.
Managed alignment and tolerance adjustments across all subsystems to ensure seamless integration between structural and moving parts.
Electronics & Architecture
Directed the high-level design of the electrical architecture, defining voltage levels, power separation, and subsystem coordination.
Selected and sourced key components such as:
Motors (hub and BLDC)
ODrive motor controllers
Jetson Orin Nano
Component Sourcing & BOM Management
Solely responsible for sourcing all mechanical and electrical components, including bearings, aluminum extrusions, fasteners, gear parts, and sensors.
Managed bill of materials (BOM) logistics, ensuring on-time procurement and compatibility across subsystems.
Simulation & Software Exploration
Conducted preliminary research into ROS2 simulation environments, including control frameworks and package integration for bipedal robots.
While implementation was not successful due to time constraints, the investigation helped inform the system’s future scalability and ROS2 integration path.