Quantum Computing for Self-Driving Cars and Pedestrian Detection

Abstract

Quantum computing and self-driving cars are at the forefront of modern-day technological progress. One of the most crucial tasks for the safety of self-driving cars is pedestrian detection, which must be quick and accurate. Due to the enormous amounts of data involved, quantum computers are ideal as they provide exponential speedup. The benefits of quantum are twofold. First, the speed and accuracy of object detection is significantly improved. Second, groups of self-driving cars can communicate effectively with a centralized quantum computer, which leads to path selection efficiency. This paper presents a novel implementation of a k nearest neighbors classification algorithm, in both classical and quantum versions. The experimental procedure is outlined in four steps: data collection, classical nearest neighbor model, quantum nearest neighbors, and a wireless networking framework. The data consists of square grayscale images from the PennFudan (Shi, 2007) and DaimlerMono (Gavrila., 2012) datasets. Both models were tested by running them on testing data independent of training data. The classical model achieved fifty percent accuracy rate, and the quantum model ninety five percent. Due to the unavailability of real quantum computers, runtime was not tested, but theoretical constructs estimate the speedup to be on the order of millions for moderate input sizes. The networking framework was qualitatively analyzed to be feasible and sufficient. The successful implementation of pedestrian detection serves as a proof of concept and can be extended to a broader range of detections. The results demonstrate quantum computers as a possible solution to safe self-driving car technology.