Robot navigation is the difficulty of guiding a robot to its intended destination. Autonomous navigation is contingent upon motion planning, i.e., the interdependence between activities such as environment sensing, path planning, and path tracking, among others. This research paper proposes and develops a four-wheel differential drive for the autonomous rover. The rover used computer vision techniques to avoid a collision with obstacles and walls and predict the Rover's direction of motion, a Global Positioning System (GPS) and magnetometer sensor were used to reach its target location. Transfer learning was used to retrain the ResNet-18 model to our dataset. A camera was mounted in front of the rover to capture the images, and a powerful low-cost embedded computer Nvidia Jetson Nano was used to process those images and predict the Rovers direction of motion. A GPS sensor was used to get the rover’s current location, and a magnetometer was used to get the rover’s current heading from which the distance from the target location and required heading were calculated. PID controller was used to calculate the speed of motors at which the rover will steer and move towards the target location. A functional autonomous Rover was successfully developed with computer vision techniques to avoid obstacles. Many possibilities exist to improve the model proposed in this paper. LIDAR can enable a 360 obstacle detection functionality, which could allow it to avoid obstacles from all directions.