top of page
Daniel Vinals Robotics and Mechatronics Engineering Portfolio
Welcome to Daniel Vinals' Engineering Portfolio
.jpg)
Greetings! I am an interdisciplinary robotics and mechatronics engineer with a passion for the convergence of AI and hardware development. My experience spans diverse engineering sectors, including medical devices, drone technology, advanced manufacturing, and fulfillment robotics.
As a Master's candidate in Robotic and Autonomous Systems at Boston University and an Amazon Robotics Day One Fellow, I've had the opportunity to work on cutting-edge projects across various industries. From developing AI-guided medical devices at MIT Lincoln Lab to designing advanced robotic arm systems for warehouse automation at Amazon Robotics, I've consistently pushed the boundaries of what's possible in intelligent hardware systems.
My portfolio includes a range of school projects that demonstrate my skills in mechatronics, robotics, and medical device engineering. These projects showcase my ability to apply theoretical knowledge to practical challenges, resulting in innovative solutions in areas such as autonomous systems, smart medical devices, and advanced robotic applications. Each project reflects my commitment to blending engineering principles with cutting-edge control techniques to create intelligent and efficient hardware solutions.
I invite you to explore these academic endeavors and see how my interdisciplinary approach and diverse skill set translate into promising innovations in the fields of robotics, mechatronics, and medical technology.
Let's connect and discuss how we can shape the future of intelligent hardware systems!
My Portfolio
Welcome to my portfolio. Here you’ll find a selection of my work. Explore my projects to learn more about what I do.
Robotic Arm Work Cell with Force-Sensing Closed-Loop Control
Amazon Robotics wanted to automate the task of placing items from totes into bags and boxes — a high-volume fulfillment step with the potential to save billions through automation. The main bottleneck was that vision-only systems couldn't reliably prevent damage: relying on perception alone, the arm had no way to sense contact during placement, risking harm to the robot, the packaging enclosure, or the picked object itself. I led full-stack development of a work cell that closed this gap by adding force sensing to the loop, taking the system from concept through integration and testing on real hardware.
At the core of the cell is a UR10 robotic arm paired with Zivid point cloud cameras, a Robotiq force/torque sensor, and closed-loop controllers. I designed and fabricated a custom suction-based end effector that integrated mechanical, pneumatic, and electrical subsystems, and developed the eye-to-hand calibration between the end effector and the cameras so the arm could act accurately on what it saw.
For perception, I built a computer-vision program using YOLO, OpenCV, and the Zivid cameras to recognize objects and enclosures in cluttered totes and compute their 3D poses relative to the arm's base. The controls side is where the system really came together: I implemented a behavior-tree force controller with closed-loop feedback for placement operations, backed by a dynamic end-effector model with force-sensor compensation calibration. Rather than relying on position alone, the arm senses contact forces during placement and adjusts in real time — letting it set objects down gently and reliably despite variation in object size, weight, and pose, without damaging the object, the enclosure, or itself.
At the core of the cell is a UR10 robotic arm paired with Zivid point cloud cameras, a Robotiq force/torque sensor, and closed-loop controllers. I designed and fabricated a custom suction-based end effector that integrated mechanical, pneumatic, and electrical subsystems, and developed the eye-to-hand calibration between the end effector and the cameras so the arm could act accurately on what it saw.
For perception, I built a computer-vision program using YOLO, OpenCV, and the Zivid cameras to recognize objects and enclosures in cluttered totes and compute their 3D poses relative to the arm's base. The controls side is where the system really came together: I implemented a behavior-tree force controller with closed-loop feedback for placement operations, backed by a dynamic end-effector model with force-sensor compensation calibration. Rather than relying on position alone, the arm senses contact forces during placement and adjusts in real time — letting it set objects down gently and reliably despite variation in object size, weight, and pose, without damaging the object, the enclosure, or itself.
Multi-Arm Robotic System for Large Object Orientation Control
I developed this multi-arm robotic manipulation system as my master's degree independent study project, working closely with advisor Professor Roberto Tron over a full semester.
The Challenge: Traditional methods for manipulating very large objects in manufacturing, construction, and aerospace lack the precision, flexibility, and safety needed for complex tasks. Individual robotic arms are limited by their workspace constraints when handling objects significantly larger than their reach.
My Solution: I created a novel "gait-inspired" control strategy that coordinates multiple robotic arms to reorient large objects through a walking-like pattern:
Stepping Phase: Individual robots strategically reposition their suction-based contact points
Dragging Phase: All robots work together to apply coordinated forces for object reorientation
Results: Using PyBullet simulation with four UR5e robotic arms, I successfully demonstrated reorientation of objects up to 5×5×1 meters (approximately 5x larger than any single robot's workspace) with rotations up to 100 degrees in roll and pitch axes.
The Challenge: Traditional methods for manipulating very large objects in manufacturing, construction, and aerospace lack the precision, flexibility, and safety needed for complex tasks. Individual robotic arms are limited by their workspace constraints when handling objects significantly larger than their reach.
My Solution: I created a novel "gait-inspired" control strategy that coordinates multiple robotic arms to reorient large objects through a walking-like pattern:
Stepping Phase: Individual robots strategically reposition their suction-based contact points
Dragging Phase: All robots work together to apply coordinated forces for object reorientation
Results: Using PyBullet simulation with four UR5e robotic arms, I successfully demonstrated reorientation of objects up to 5×5×1 meters (approximately 5x larger than any single robot's workspace) with rotations up to 100 degrees in roll and pitch axes.
MIT Robotic Surgical Device
During my time at MIT LL, my primary functions were to design, fabricate, and test patented R&D hardware for AI-GUIDE (Artificial Intelligence-Guided Ultrasound Interventional Device), which is a semi-autonomous medical device in development for the Department of Defense.
More info about the project can be found here: https://www.ll.mit.edu/r-d/projects/artificial-intelligence-guided-ultrasound-intervention-device
Due to the top-secret nature of the DOD, I cannot disclose everything that I did, but I can speak generally of what my projects entailed, as well as describe the projects that are currently visible to the public.
I was solely responsible for developing the cartridge components used in AI-Guide's semi-autonomous tracheotomy procedure. My work encompassed the entire product development cycle: 3D modeling the cartridge designs, overseeing manufacturing processes, and conducting comprehensive testing protocols. I independently designed and fabricated custom testing assemblies incorporating 3D printed fixtures and force measurement instrumentation to enable iterative, data-driven design optimization and validate requirement fulfillment. The team is currently filing for a patent on my design.
Other project contributions included:
Design for Manufacturing: Optimized part designs for vacuum forming and injection molding processes, considering manufacturing constraints and cost efficiency
Custom Tooling: Designed and fabricated a 3D printed vacuum forming setup for in-house prototyping capabilities
Performance Validation: Created standardized testing procedures and built specialized test rigs to measure and quantify hardware component performance metrics
More info about the project can be found here: https://www.ll.mit.edu/r-d/projects/artificial-intelligence-guided-ultrasound-intervention-device
Due to the top-secret nature of the DOD, I cannot disclose everything that I did, but I can speak generally of what my projects entailed, as well as describe the projects that are currently visible to the public.
I was solely responsible for developing the cartridge components used in AI-Guide's semi-autonomous tracheotomy procedure. My work encompassed the entire product development cycle: 3D modeling the cartridge designs, overseeing manufacturing processes, and conducting comprehensive testing protocols. I independently designed and fabricated custom testing assemblies incorporating 3D printed fixtures and force measurement instrumentation to enable iterative, data-driven design optimization and validate requirement fulfillment. The team is currently filing for a patent on my design.
Other project contributions included:
Design for Manufacturing: Optimized part designs for vacuum forming and injection molding processes, considering manufacturing constraints and cost efficiency
Custom Tooling: Designed and fabricated a 3D printed vacuum forming setup for in-house prototyping capabilities
Performance Validation: Created standardized testing procedures and built specialized test rigs to measure and quantify hardware component performance metrics
Cube Stacking Robot
The goal for the stacking machine was for it to function like an arcade machine. It was important to us that anyone could easily pick up the controls and build a tower without needing to write any code. We started with an Arduino Mega running the class provided firmware that allowed it to run G-code. Using Repetier, a software for controlling 3D printers, we sent G-code commands to the machine through the Arduino's serial connection.
Our design consisted of the usage of four stepper motors and a servo to produce 3.5 degrees of freedom. One stepper motor was designated to move the end effector from left to right. Two stepper motors were designated to move the previous setup mentioned up and down. A belt drive system was used to achieve the desired linear motion. Acrylic and T-slotted aluminum extrusions (8020) were used to make the railings and carriages that produced the linear movement. The last stepper motor was used to rotate the cube holder ring. This was achieved by using an internal gear and a spur gear controlled by the motor. This setup was made from laser-cut acrylic. The end-effector (gripper) was driven by two meshing, equal number teeth gears. These two gears powered two four bar linkage mechanisms that opened and closed the fingers of the claw. The servo motor was connected to the input link allowing the two gripper arms to open and close based on the position angle of the servo.
By default, G-code used absolute positioning. This meant each movement command had to tell the machine what position to move to relative to zero. Because we wanted to use a controller with simple button inputs, it was necessary to use relative positioning, which allowed programming the machine to move a specified distance in a certain direction.
Once we had the G-code, we used reWASD to create macros that mapped the commands onto the buttons of an Xbox controller. With the controller connected to the laptop via bluetooth, we could control all 3.5 degrees of freedom wirelessly simply by triggering the movement commands with button presses. After some trial and error, we modified the movement distances in order to strike a balance between precision and ease of use. Smaller increments offered more control, but increased the number of times the user had to press a button to move in a certain direction. We settled on 5 mm for left/right movement, 10 mm for up/down movement and 5 degrees for the turntable. We found these values allowed for easy yet precise control of the stacker.
Our design consisted of the usage of four stepper motors and a servo to produce 3.5 degrees of freedom. One stepper motor was designated to move the end effector from left to right. Two stepper motors were designated to move the previous setup mentioned up and down. A belt drive system was used to achieve the desired linear motion. Acrylic and T-slotted aluminum extrusions (8020) were used to make the railings and carriages that produced the linear movement. The last stepper motor was used to rotate the cube holder ring. This was achieved by using an internal gear and a spur gear controlled by the motor. This setup was made from laser-cut acrylic. The end-effector (gripper) was driven by two meshing, equal number teeth gears. These two gears powered two four bar linkage mechanisms that opened and closed the fingers of the claw. The servo motor was connected to the input link allowing the two gripper arms to open and close based on the position angle of the servo.
By default, G-code used absolute positioning. This meant each movement command had to tell the machine what position to move to relative to zero. Because we wanted to use a controller with simple button inputs, it was necessary to use relative positioning, which allowed programming the machine to move a specified distance in a certain direction.
Once we had the G-code, we used reWASD to create macros that mapped the commands onto the buttons of an Xbox controller. With the controller connected to the laptop via bluetooth, we could control all 3.5 degrees of freedom wirelessly simply by triggering the movement commands with button presses. After some trial and error, we modified the movement distances in order to strike a balance between precision and ease of use. Smaller increments offered more control, but increased the number of times the user had to press a button to move in a certain direction. We settled on 5 mm for left/right movement, 10 mm for up/down movement and 5 degrees for the turntable. We found these values allowed for easy yet precise control of the stacker.
Autonomous Charger Robot for Electric Vehicles
This project conducted a comprehensive analysis of an innovative low-cost robotic platform engineered for autonomous electric vehicle charging, driven by the imperative to automate farming practices. Key features including belt-driven actuators, prismatic and gimbal-type joints, and a reflective platform were explained in detail, illuminating the mechanical design and actuation system. The integration of timing pulleys, belt tensioners, stepper motors, and encoders facilitated precise control over the platform's movements. Moreover, emphasis was placed on the development of an advanced inverse kinematics model, pivotal for achieving precise manipulation of the platform's end effector. To illustrate the platform's capabilities in end effector pose control, a Python simulation was conducted. Ultimately, these efforts contributed to the development of autonomous technologies, poised to address contemporary agricultural challenges and beyond.
Wearable, Vibrotactile Sensory Substitution Device for Balance Rehabilitation
Our project aimed to create a wearable, assistive rehabilitation device for people with sensory impairment in the legs as a result of SCI, Stroke, MS, Diabetes, or other similar conditions. More specifically, we aimed to target patients with nerve damage in the sacral spine but not in the lumbar region. Sacral spine nerve damage could result in loss of force sensation in the bottom of the foot, impairing patients' balance. However, by redirecting force feedback to lumbar nerves that end in the top of the foot, our device was designed to restore balance via sensory substitution and neuroplasticity.
Research had been conducted into various methods of balance restoration, through various means. Various forms of feedback had been used, including visual and vibrotactile. Likewise, research had been conducted into sole sensing for gait monitoring. Our device was intended to be a novel combination of the two, allowing for both balance detection and sensory substitution in a single, compact package.
The device was designed to have three main components. Sensing was to take place in an insole containing an array of force sensing resistors (FSRs). An Arduino Nano, housed in an enclosure strapped to the patient's ankle, would read and interpret the real-time spatially-distributed weight data and control an array of piezo vibration motors mounted to the top of a wearable sock. To avoid excess computation by the Arduino board, the piezo vibrators were designed to share the same arrangement as the FSRs under the foot, and each FSR would have a corresponding piezo actuator.
By relaying underfoot force data to the top of the foot using vibrotactile feedback, we hoped to intuitively restore the patient's sense of weight distribution. As more force was applied to a given FSR, the magnitude of felt vibration would increase. The result was expected to improve gait, posture, and balance, better enabling patients' activities of daily life while reducing the risk of falling.
Research had been conducted into various methods of balance restoration, through various means. Various forms of feedback had been used, including visual and vibrotactile. Likewise, research had been conducted into sole sensing for gait monitoring. Our device was intended to be a novel combination of the two, allowing for both balance detection and sensory substitution in a single, compact package.
The device was designed to have three main components. Sensing was to take place in an insole containing an array of force sensing resistors (FSRs). An Arduino Nano, housed in an enclosure strapped to the patient's ankle, would read and interpret the real-time spatially-distributed weight data and control an array of piezo vibration motors mounted to the top of a wearable sock. To avoid excess computation by the Arduino board, the piezo vibrators were designed to share the same arrangement as the FSRs under the foot, and each FSR would have a corresponding piezo actuator.
By relaying underfoot force data to the top of the foot using vibrotactile feedback, we hoped to intuitively restore the patient's sense of weight distribution. As more force was applied to a given FSR, the magnitude of felt vibration would increase. The result was expected to improve gait, posture, and balance, better enabling patients' activities of daily life while reducing the risk of falling.
Smart Pipe Cutting Machine
A machine that can autonomously cut freshly extruded pipes of various diameters and lengths.
-Developed a CAD model of the entire machine
-Calculated and specified actuator and linear drive requirements
-Developed a financial justification for the value of the machine
-Developed a CAD model of the entire machine
-Calculated and specified actuator and linear drive requirements
-Developed a financial justification for the value of the machine
Abnormality Detection in Musculoskeletal Radiographs using a Neural Network Model
Our project aimed to design an advanced automated detection system for identifying anomalies in musculoskeletal radiographs. Inspired by the concept discussed in a referenced paper, this system sought to detect fractures, lesions, and joint abnormalities with high precision. The primary challenge lay in detecting subtle anomalies within intricate anatomical structures, which could be compounded by variations in imaging techniques and individual patient anatomy. Another challenge was the different types of abnormalities and different body parts. For this project, we aimed to create a robust system that could effectively detect if a radiograph image was abnormal, thus supporting radiologists by improving diagnostic accuracy and efficiency.
Vertical Farming Harvesting Robot
The aim of this project was to not only develop an optimal warehouse shelving plan that would allow for the greatest possible usage of energy and space while reducing costs, but to also develop a fully automated robot that could harvest the plants without the need for transporting cells or having workers manually care take. This prototype was only constrained to harvesting strawberries and did not feature any additional systems, such as water and light systems.
The harvesting robot was designed to recognize strawberries using computer vision technology and harvest them via a robotic arm with 6 degrees of freedom. The prototype also included a scalable shelving architecture with troughs for the strawberries and a railing system for the harvesting robot to maneuver across.
The robot was designed to detect a ripe strawberry along the trough via a depth camera and computer vision. This would signal the linear drives to move the robot to the region of the strawberry and then have the arm move to the location of the strawberry. The claw (end effector) was designed to cut the stem of the strawberry, which would be pulled towards the claw via a vacuum pipe connected to it, and would fall into the vacuum pipe leading it to a storage basket that was below the robot.
The harvesting robot was designed to recognize strawberries using computer vision technology and harvest them via a robotic arm with 6 degrees of freedom. The prototype also included a scalable shelving architecture with troughs for the strawberries and a railing system for the harvesting robot to maneuver across.
The robot was designed to detect a ripe strawberry along the trough via a depth camera and computer vision. This would signal the linear drives to move the robot to the region of the strawberry and then have the arm move to the location of the strawberry. The claw (end effector) was designed to cut the stem of the strawberry, which would be pulled towards the claw via a vacuum pipe connected to it, and would fall into the vacuum pipe leading it to a storage basket that was below the robot.
bottom of page