Autonomous Robot Car

Building a robot that moves is relatively easy. Building one that can decide where it should move next is much harder.

I built RobotCar as an experimental platform for learning how autonomous machines connect sensing, software, and physical movement. The car uses a Raspberry Pi to control four motors while three ultrasonic sensors measure the space in front of and beside it. The software uses those measurements to decide whether to continue forward, stop, reverse, or turn toward a clearer path.

The project also includes manual keyboard control, which lets me test the drivetrain directly and temporarily take control even while autonomous mode is running.

RobotCar combines a small Linux computer with motor-control electronics and distance sensors.

The combination turns the car into a physical computing system where software decisions immediately become real movement.

The first autonomous version uses a straightforward sense → decide → move loop.

1. DriveThe car moves forward while checking its left, center, and right distance sensors.
2. DetectIf an obstacle becomes too close, the car stops instead of continuing blindly.
3. Create spaceIt can reverse to give itself room to maneuver.
4. CompareThe software compares the available space around the car.
5. ChooseIt turns toward the clearer direction.
6. ContinueOnce the new path is selected, the car resumes driving and begins sensing again.

This creates basic autonomous obstacle avoidance without requiring a remote driver. A manual override is also available while autonomous mode is active.

Before testing autonomy, I built keyboard controls so each movement could be tested independently. Manual driving became an important debugging tool: it separates motor and wiring problems from navigation-logic problems.

Watch a short demonstration of the vehicle moving and responding to its environment.

If the embedded player does not load, watch the demonstration on YouTube ↗.

The repository includes separate programs for motor testing, ultrasonic-sensor testing, manual driving, and autonomous driving, along with build, wiring, setup, and parts documentation.

Distance sensors can tell the robot that something is nearby, but they cannot tell it what that object is or whether an open-looking area is actually safe to drive across.

The next stage is camera-based navigation: identifying driveable floor and making better path decisions using visual information rather than relying entirely on three distance measurements. That moves the project from reactive obstacle avoidance toward a more complete autonomous-navigation system.

RobotCar taught me that autonomy is not one piece of code. It is a chain of systems that all have to work together.

• A navigation algorithm can be correct while a motor is wired backward.
• A sensor can produce measurements while still giving unreliable information to the decision logic.
• I had to transition from the L298N to the DRV8833 because power loss in the L298N reduced the car's maneuverability and control. The DRV8833 provided a more efficient motor-control solution for the autonomous system.
• Building the car taught me to debug at the boundary between hardware and software, where changing a wire, a sensor position, or a few lines of Python can completely change how the physical system behaves.