Hand-Gestured Stroller

Embedded control with TM4C123 and an MPU6050 accelerometer for hands-free stroller navigation.

Project overview. The Hand-Gestured Stroller is an embedded systems project designed to alleviate the physical strain of manually pushing baby strollers. Using a Tiva C Series TM4C123 microcontroller and an MPU6050 accelerometer, this project creates a vehicle that autonomously moves, turns and stops based on the user’s hand orientation in real time.

Introduction

Parenting often involves significant physical exertion, particularly when transporting toddlers. While the baby stroller (invented in 1733) solved the issue of carrying a child, it still requires manual pushing. This becomes increasingly difficult for parents with multiple children or physical limitations.

Our team developed the Hand-Gestured Stroller. Instead of a physical handle, the user controls the stroller via a handheld module containing an accelerometer. The system interprets hand tilt to drive the motors, offering a hands-free experience.

System architecture

1. Hardware configuration

  • Master controller: TM4C123GXL (ARM Cortex-M4)
  • Sensor: MPU6050 (3-axis accelerometer + gyroscope)
  • Actuation: 2× Noyito 170 W high-power H-bridge motor drivers
  • Power: 2× 12 V 6 A batteries (series for 24 V)
  • Debugging: Arduino Uno (bridged via UART for serial monitoring)

Hardware schematic diagram

Figure 1: Schematic showing connections between the TM4C123, MPU6050 and H-bridge drivers.

2. Communication protocols

  1. I2C — between the TM4C123 and MPU6050. The microcontroller reads the X/Y/Z acceleration registers at 1 kHz.
  2. PWM — controls the speed and direction of the DC motors via the H-bridges.

Methodology

1. Finite state machine

The control logic is implemented as an FSM. The accelerometer outputs raw data ranging 0g–4g. We map specific threshold ranges to five distinct states: STOP, DRIVE, REVERSE, LEFT, RIGHT.

\[\text{State} = \begin{cases} \text{DRIVE} & \text{if } X \ge 3.5,\ Y \le 0.5,\ 0.5 \le Z \le 1.5 \\ \text{LEFT} & \text{if } X \ge 3.5,\ 2.5 \le Y \le 3.5,\ Z \le 0.5 \\ \text{RIGHT} & \text{if } X \ge 3.5,\ 0.5 \le Y \le 1.5,\ Z \ge 3.5 \\ \text{REVERSE} & \text{if } 2.5 \le X \le 3.5,\ Y \le 0.5,\ Z \ge 3.5 \\ \text{STOP} & \text{Default / Out of range} \end{cases}\]

Any reading outside these “safe zones” defaults the system to STOP for safety.

Finite state machine diagram

Figure 2: FSM determining motor output based on sensor input.

2. PWM calculation

We configured a 50 Hz PWM frequency. The load value for the countdown timer is

\[\text{PWM}_{\text{LOAD}} = \frac{\text{ClockRate}}{\text{Divisor} \times \text{Frequency}} - 1\]

With a 16 MHz clock and the default divisor, this lets us set precise duty cycles for differential steering and straight-line driving.

3. Debugging with UART

The TM4C123 has no native serial monitor for runtime debugging, so we bridged TX→RX to an Arduino Uno and used the Arduino IDE Serial Monitor to troubleshoot sensor noise and logic errors.

Results

Prototype performance

The final prototype successfully:

  1. Initializes I2C and calibrates the MPU6050.
  2. Reads hand tilt data in real time.
  3. Drives the 12 V motors forward, backward and turns based on gestures.

Final prototype build

Figure 3: The assembled Hand-Gestured Stroller prototype.

Demo

Power analysis

Under load, the motors drew approximately 12 A, which caused thermal inefficiency and fast battery drain — largely due to the direct-drive configuration lacking gear reduction.

Key findings

  1. Mechanical efficiency vs. electronic control. No amount of software optimization can fix a mechanical mismatch — direct-drive motors without gearing required excessive torque, resulting in 12 A current draw and heat. A gearbox would have allowed efficient RPM with the necessary torque.
  2. Sampling-rate management. A 1 ms blind cycle delay was added after every read/write; processing data faster than the sensor could provide it caused bus errors.
  3. Cross-platform debugging. Using an Arduino as a display interface for a more powerful ARM Cortex controller is a cost-effective way to visualize data without expensive JTAG debuggers.

Conclusion

The Hand-Gestured Stroller successfully proves the concept of controlling a heavy-load vehicle using MEMS accelerometers and embedded logic. Future iterations would integrate gear reduction to lower current draw and add a PID controller to smooth jerky transitions between FSM states.

Team Members: Tony Tran, Peter Kieu, Stuart Alfafara, Zachary Nguyen.