For the new industrialization, automation is an uprising field which is evolving everyday. This can also be noticed in the Transportation industry, which is moving towards self-driving vehicles. But complete self-driving vehicles are not yet achieved. To achieve a self-driving vehicle, algorithms like lane detection, obstacle detection, traffic sign recognition etc.., can be used. If a full fledged self-driving vehicle is achieved, there will be less potential for road accidents. This paper explores both hardware and software elements of self-driving vehicles that includes the use of deep learning technique, Convolutional Neural Networks (CNN) for lane detection. Software components such as Udacity simulator, along with hardware components like raspberry Pi 3b+, pi Camera, and ultrasonic sensors, are used. In this paper, raspberry pi 3B+ is used as a controller for the four wheel rover. Images from the pi camera are delivered to the raspberry pi. A trained CNN model will forecast the steering angle. CNN algorithm will give steering direction as output by taking images from pi camera as input. This model differs from Nividia’s model in its complexity, a large parameter count, training data dependency, and algorithmic flexibility. Another key aspect is that this model’s output is the mean value of the steering value generated from Pixel Summation and the steering value generated by Nvidia’s CNN model. It provides the final output to the motor drivers, causing the motor drivers to deliver appropriate signals to the relevant motors. The outcome will be a self-driving rover which will follow an instructed path and be able to detect an obstacle if present and will be able to continue on the lane.
CNN based Curved Path Detection and Obstacle Avoidance for Autonomous Rover
17.12.2023
1754849 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Path tracking, obstacle avoidance and dead reckoning by an autonomous planetary rover
Kraftfahrwesen | 1998
|OBSTACLE AVOIDANCE, PATH PLANNING AND CONTROL FOR AUTONOMOUS VEHICLES
British Library Conference Proceedings | 2019
|