.png.png-68x68x2.png?_=Dlz52z3lMPpZDXjdwOkBpg==)
From Concept to First Mechanical Design
In my previous post, I introduced Tiltie, my entry for the element14 My Project Is Trash Design Challenge, and explained how I arrived at the Smart Tilting Bin concept.
Tiltie is an AI-powered recycling bin that uses a camera, microphone, and impact sensor to identify waste items and sort them into four bins using a motorized tilting tray. Using hand gestures detected by an infrared sensor, users can generate new training samples to improve the AI models over time or simply play a game. Built mainly from reused materials, it features a fully parametric design and educational documentation.
With the challenge kit delayed in customs, I decided to focus on what is probably the most challenging part for me: the mechanical design.
Before diving into sensors, AI, and electronics, I wanted to make sure I had a mechanism capable of moving an object from a single entry point to any of the four collection bins.
It sounded simple enough. As usual, things became a bit more interesting once I started sketching ideas.
Letting Gravity Do the Work
If you read my previous post, you may remember that the recycling-plant crane concept was getting bogged down by the gripping system. Handling objects of different shapes and sizes added a lot of complexity.
Then I realized I could avoid the problem entirely. Instead of picking objects up, why not place them on a tray and let gravity do the work?
That idea transformed the project into something much better suited for a home or small office environment.
This became the foundation of Tiltie's mechanical design. Instead of a robotic arm or a complex sorting system, it uses a circular concave tray that can rotate and tilt.
The tray has two openings on opposite sides. By adjusting its position, an object can slide through one of them and fall into the appropriate collection bin.
The concept is straightforward, although there are still plenty of details to work out.
Two Motions, Four Destinations
The core idea is based on combining two independent motions:
- Rotation: the tray can rotate 90 degrees around a vertical axis.
- Tilt: the tray can tilt up to 45 degrees in either direction around a horizontal axis.
To keep things simple, I chose a different motor for each task. Rotation is handled by a 28BYJ-48 stepper motor driven by a ULN2003 board, providing accurate positioning of the tray. Two limit switches define the travel range and establish the home position.
Tilting is performed by either an SG90 or an EMAX ES08MA II servo. Its range is more than enough to move the tray between the neutral position and the two discharge angles.
The operating sequence is straightforward. If the target bin is already aligned, the tray simply tilts. Otherwise, the assembly rotates 90° before tilting.
With just two small motors, the system can reach four sorting destinations. At 0°, waste is directed to the Paper or General Waste bins. After a 90° rotation, the same tilt motion directs items to the Packaging or Organic Waste bins.
What I particularly like about this arrangement is that it avoids the need for a turntable beneath the tray, leaving a completely unobstructed path from the input to the collection bins.
Designing the Mechanism
I then moved on to modelling the mechanism in OpenSCAD.
The mechanism consists of just a few relatively simple elements:
The Base
A fixed structure supports both the rotation mechanism and the stepper motor.
I deliberately avoided placing a solid platform beneath the tray, allowing objects to fall directly into the bins without obstruction.
The Rotating Yoke
A U-shaped structure mounted on the stepper motor shaft supports the entire tilting assembly. It rotates between 0° and 90°, orienting the tray toward one pair of bins or the other.
The Tilt Axis
A horizontal shaft supports the tray. One end acts as a simple pivot, while the other is connected to a small servo motor.
As the servo moves, the shaft tilts the tray in either direction.
The Tray
The tray is a circular concave dish approximately 150 mm in diameter, featuring two rounded openings on opposite sides of its rim.
Its shape is designed to keep objects reasonably centered while level and allow them to slide out smoothly when tilted.
The Counterweight
A small counterweight mounted beneath the horizontal shaft helps balance the tray and reduce the load on the servo motor.
The individual parts are simple; fitting them into a compact and reliable mechanism is the real challenge.
Eyes, Ears, and Gestures: Sensor Placement
The mechanical design is not just about moving the tray. It also needs to provide suitable mounting points for the sensors that will give Tiltie its "intelligence."
I wanted the camera to have a clear view of the entire tray, regardless of its position.
The simplest solution was to mount it above the mechanism, pointing downward toward the center of the tray.
To achieve this, I designed an L-shaped bracket, although the CAD model also includes an alternative arch-shaped version. Both attach to a diagonal structural member and provide a stable mounting point for the camera.
The selected camera is a Logitech C170, connected to a Raspberry Pi CM4. It captures an image of each object as it arrives on the tray, while its built-in microphone records the impact sound. Combined with accelerometer data, these signals should enable more reliable object classification than any single source alone.
The gesture sensor requires a slightly different approach. The challenge kit includes the Analog Devices EVAL-CN0569-PMDZ module, based on the company's infrared light angle sensing technology. This sensor will be mounted beneath the front panel, facing the tray.
Keeping Things Balanced
The counterweight reduces the load on the servo and helps the tray return to its horizontal position. Since the required balance is not yet known, both the counterweight position and arm length are adjustable and will be tuned during assembly.
Finding the Right Mechanical Layout
I explored two different configurations: a wider yoke with the support arms outside the tray, and a more compact version with the arms beneath it and the servo mounted on the tilting assembly.
Both variants coexist in the same OpenSCAD model, making them easy to compare and refine. For now, I am focusing on the compact version, although several details still need validation once the mechanism is physically assembled.
Homing and Position Control
The rotation mechanism needs to determine its two end positions. To achieve this, I am using two CamdenBoss CSM40500A microswitches, one at each end of the 90° travel range.
A cam mounted on the rotating yoke actuates the switches, providing the reference points required for homing. The switches are mounted in adjustable slots so their trigger positions can be tuned during assembly. The final cam geometry will be validated once the physical mechanism has been built.
A Fully Parametric Design
I chose OpenSCAD because a code-based workflow feels natural to me. The entire design is parametric, allowing key dimensions to be adjusted and the assembly regenerated in seconds.
This flexibility supports one of my project's goals: reusing household waste wherever possible instead of relying on specific materials.
A plastic container, a curtain rod tube, or a scrap piece of wood could easily be incorporated into the design.
Bringing It to Life: The First Simulation
With the mechanical layout in place, I added a simple animation to the OpenSCAD model.
It reproduces the complete sorting sequence for all four waste categories, verifying that the tray can rotate, tilt, and reach every bin without interference. Although it is not a physics-based simulation, it provides a quick way to validate the design before printing any parts.
What's Next?
The parametric model and basic sorting simulation are now complete, but the real test comes next.
The immediate goal is to build the mechanism using a mix of printed and recycled parts, verify clearances and balance, and integrate the 28BYJ-48 stepper motor, ULN2003 driver board, SG90 servo, and homing switches.
Once the mechanics are working reliably, I can focus on the electronics and software, combining camera, audio, and ADXL372 accelerometer data to support object classification, gesture recognition, and game mode features.
Tiltie is still far from being a finished smart bin, but it is finally starting to take physical shape.
Thanks for reading, and stay tuned for the next update!