Haptic Matrix - A modular vibrational haptic grid

2024-2025 / built with my team through the Tufts Idea Lab

Summary

The wired haptic grid on the workbench next to the soldering station — [1]The prototype grid on the bench, every motor channel hand-wired.

I was invited to build out the technical work and hardware for a vibration grid: a soft TPU mesh_[1] carrying a 5×5 matrix of individually addressable vibration motors, with sensors embedded in the material alongside them. Each cue lands in about 5 ms, fast enough to keep touch synchronized with what you see and hear. The motors are PWM-driven from custom PCBs, patterns can be customized remotely over Wi-Fi or Bluetooth, and the flexibility comes from the mesh itself rather than the boards.

Approach

3D render of the custom driver PCB with the 5x5 motor array and ESP32 module — [2]The custom driver board: 25 motor channels arrayed around an ESP32-WROOM-32.

The device treats touch the way a display treats pixels: each of the 25 motors in the grid is individually addressable, so the system renders localized tactile patterns the same way an LED matrix renders an image. An ESP32-WROOM-32 module drives the grid through custom PCBs_[2], and I2C GPIO expanders scale the channel count while keeping latency low_[3]. Each cue lands in 5 ms, which keeps tactile output synchronized with visual and auditory feedback in real time. The motors are PWM-controlled with optimized power distribution, and Wi-Fi and Bluetooth support remote customization of the patterns.

PCB copper layout of the motor driver array — [3]The layout: 25 driver channels fanned out from the I2C expanders.

The body is a soft mesh: vibration motors embedded on a grid within TPU, with sensors embedded in the material alongside them. We experimented with different internal geometries of the TPU to tune how the mesh flexes. Flexibility comes from the material, not the boards. We explored flex PCBs, but SMT vibration motors were not the best fit for the project, and the motors had to be wired in regardless.

The target uses were targeted pain relief, accessibility, and interactive applications: therapy, gaming, and human-computer interaction.

Design Process

PCB components laid out on a table, ready for assembly — [4]Parts arriving from the fab and distributors, laid out for assembly.

I built the device with my team through the Tufts Idea Lab over a semester. The work:

Custom PCBs_[4], designed and assembled by hand, refining motor control and power efficiency.
A 5×5 LED prototype grid to test haptic-visual interaction.
Teaching: PCB design and reflow soldering for the other team members.
Soft-robotics experiments with actuator-driven designs for higher tactile resolution.

Applications

Therapeutic and assistive technology. Targeted vibrational therapy for pain relief and muscle relaxation; an alternative communication channel for visually impaired users; integration into physical therapy and rehabilitation programs.

Immersive media and gaming. Spatial haptic effects for VR/AR, embedding in gaming peripherals for added realism, and tactile cues for gesture-based control.

Soft robotics and adaptive interfaces. The soft-robotics experiments point toward versions of the grid built as adaptive, conformable interfaces.

Future Directions

The direction I care most about is beamforming. Because every motor in the grid is individually driven, the array can do what phased arrays do with antennas and speakers: fire elements with coordinated timing so their waves interfere, steering and focusing energy into patterns far more complex than any single motor can produce. Bartosz Ciechanowski's interactive Sound builds the interference intuition beautifully, and Jeff McBride's acoustic beamforming notes walk through building a real delay-and-sum array on a microcontroller.

The same grid works in reverse. With ultrasonic receivers embedded in the mesh, models could be trained on the returns to detect what's inside the body, imaging with increasing precision down the line. This is no longer a fringe idea: Midjourney Medical announced a full-body ultrasonic scanner built from half a million grain-of-sand-sized ultrasonic emitters that scans the body in under 60 seconds, the same beamforming-grid bet at industrial scale.

Nearer-term improvements:

Biocompatible materials for soft-robotic versions of the grid.
AI-driven real-time haptic adjustment using physiological data.
Expanded actuator options, including servos and linear resonant actuators.

I'd love to continue this idea at some point. If you have thoughts on scaling the grid, processing the sensor data, or want to collaborate, email me at daniel@integralmimetics.com.