Many industries stand to benefit from the commercialisation of quantum computing, particularly those industries that need high levels of processing power, such as the autonomous vehicle market. Quantum computers can provide a huge increase in processing speed for a number of applications in chemistry, materials science, and general linear algebra operations, and their potential for use within finance and pharmaceuticals is being explored. In this project, we will explore and develop quantum computing solutions for autonomous vehicles, and more specifically driverless cars. The aim of this project is to develop an end-to-end control system deployed in cars, where quantum computers are used to enhance the decision-making process in the control system. Autonomous systems need to repeatedly take decisions as to whether they should take a specific action or not. This is a difficult challenge, particularly when the input from different sensor data is considered. For example, deciding whether a lane change is safe is relatively straight forward for humans, but is difficult for automated control systems. QCs process data in an inherently parallel way, with a possibilistic outcome of the measurements. These can provide complementary information to the control system and hence enhance its decision-making capabilities. In a recent joint research collaboration, Massive Analytic Limited (MAL) and the National Physical Laboratory (NPL) have demonstrated that neural networks implemented on quantum computers, the so-called Quantum Neural Networks, can predict the safety of specific autonomous car manoeuvres. This result was shown on a simplified system as proof of concept. In this project, we will extend this to a real-life scenario, where the decision depends on the positions and velocities of multiple surrounding cars, and integrate the quantum neural networks in MAL's end-to-end commercial APACC control framework of a driverless car. To this aim, we will combine the expertise in control systems of MAL and the quantum software expertise at NPL, and use the autonomous systems dynamics and test facilities at the Centre for Autonomous and Cyber-Physical Systems in Cranfield University. Quantum Neural Networks have been shown to train faster than classical models for certain cases, and hence have the potential to outperform classical machine learning algorithms used in the autonomous vehicle industry. We will systematically assess this in the project. If successful, it will be a disruptive enhancement to MAL's commercial APACC control system, giving it a significant advantage over competitors.

Massive Analytic Limited will firstly build a digital twin of a section of rail track near Workington, demonstrating the availability, strength and latency of different terrestrial mobile phone signals, as well as satellite signals. We will demonstrate how, by utilising our patented AI technology to combine terrestrial and satellite signals to meet predicted demands in real-time, we can enhance on board voice and data connectivity for customers travelling by rail, as well as improve connectivity for train staff and provide better real-time information on train locations. This will be the first time that our technology has been used in the rail sector. Secondly, we will apply our technology to a Network Rail maintenance vehicle to demonstrate that the system can operate effectively in a real world setting. This will provide evidence that the system can be applied to a real world setting, enabling us to move forward with our plan to apply the technology to passenger trains in the future. Our project has been put together in collaboration with Network Rail as the first step in developing a system could be used to meet the future communication needs of railway staff and passengers.

The large-scale multimodal sensor fusion of internet of things (loT) data can be transformed into a N-dimensional classical point cloud. For example, the transformation may be the fusion of three imaging modalities of different natures such as LiDAR (light imaging, detection, and ranging), a set of RGB images, and a set of thermal images. However, it is not easy to process a point cloud because it can have millions or even hundreds of millions of points. Classical computers therefore often crash when operating a point cloud of multimodal sensor data.The emerging quantum computing technology can help users to solve the multimodal sensor point cloud processing problem more efficiently.The development of the quantum computing hardware is proceeding at a fast pace, and current quantum computers exist, with the number of quantum-bits (qubits) per computer steadily increasing. Quantum computation is therefore expected to become an important and effective tool to overcome the high real-time computational requirements. In order to operate point clouds in quantum computers, there are two problems to be solved, and these are quantum point cloud representation and quantum point cloud processing. Quantum representations of two-dimensional images abound. However, there is a distinct paucity of methods to express a three-dimensional image using quantum representation. Furthermore, to provide a quantum computing based solution for fused multimodal sensor data the representation and processing needs to be further generalized to N-dimensional quantum point clouds.We have theoretically demonstrated that representation and processing of QPCs is possible if the quantum computers have no inherent errors. Existing and near-term quantum computing hardware is noisy, so that any proposed quantum algorithm needs to be tested for its resilience to this noise. In this project we will therefore perform QPC processing also on real noisy quantum hardware. We will first simulate QPCs including noise and perform uncertainty quantification to understand its effects on QPCs. A systematic metrological comparison between CPC and QPC on noisy quantum computers will be performed. This includes definitions of measures for the efficiency and accuracy of QPC results, such as the uncertainty induced by the noisy hardware when preparing and processing the quantum point cloud, and the evaluation of the statistical variations of QPC outcomes.

Video Precognition for a Safer Network Rail Massive Analytic's Nethra:VideoAnalytic's platform optimises the use of CCTV feeds in Network Rail stations. It's AI uses deep learning to detect dangerous objects, behaviours and illegal activities alerting station staff to any issues or dangers and with it's unique augmented reality technology Nethra empowers station staff with a wholistic view of the entire station, while also allowing passengers to maintain their privacy. Nethra is the one stop solution to drastically cutting response times and enabling staff to operate efficiently for safer Network Rail stations.

The large-scale multimodal sensor fusion of internet of things (loT) data can be transformed into a N-dimensional classical point cloud. For example, the transformation may be the fusion of three imaging modalities of different natures such as LiDAR (light imaging, detection, and ranging), a set of RGB images, and a set of thermal images. However, it is not easy to process a point cloud because it can have millions or even hundreds of millions of points. Classical computers therefore often crash when operating a point cloud of multimodal sensor data. The emerging quantum computing technology can help users to solve the multimodal sensor point cloud processing problem more efficiently. The development of the quantum computing hardware is proceeding at a fast pace, and current quantum computers exist, with the number of quantum-bits (qubits) per computer steadily increasing. Quantum computation is therefore expected to become an important and effective tool to overcome the high real-time computational requirements. In order to operate point clouds in quantum computers, there are two problems to be solved, and these are quantum point cloud representation and quantum point cloud processing. Quantum representations of two-dimensional images abound. However, there is a distinct paucity of methods to express a three-dimensional image using quantum representation. Furthermore, to provide a quantum computing based solution for fused multimodal sensor data the representation and processing needs to be further generalized to N-dimensional quantum point clouds. The project will therefore involve the development and analysis of an N-dimensional quantum point cloud, and a systematic metrological comparison between CPC and QPC will be performed. This includes definitions of measures for the efficiency and accuracy of QPC results, such as the time it takes prepare and process the quantum point cloud and the evaluation of the statistical variations of QPC outcomes. The project will also evaluate how an N-dimensional quantum point cloud addresses the problem of uncertainty in multi-modal sensor data, such that precognitive/predictive models can be derived with outcomes of greater certainty than classical information processing methods.

Founded in 2010, Massive Analytic is a London-based data science company, developing innovative technologies related to artificial intelligence, big data and predictive analytics. Holding core patents for artificial precognition, MA seek to disrupt the way we interact with data, realised through the novel software platform Oscar AP. This project aims to transform video analytics by making it possible to predict likely event outcomes. Surveillance videos can be monitored automatically in real time, triggering alerts to be sent to response teams such as emergency services as required. This allows crowd managers and emergency services to target resources effectively. Current video analytic solutions are limited by their capability to capture relevant information from video streams in real time, given the increasing size of data streams. Human operators can only monitor a limited number of screens and their reliability decreases dramatically after 20 minutes continuous monitoring. The performance of automatic video analytic solutions is also limited, with high rates of false alarms and missed events. Massive Analytic have applied the latest techniques in data science and, by approaching video analytics in an entirely new way, have succeeded in demonstrating the concept of video precognition. This demonstration system is able to predict when a fight is about to break out in a crowd, or identify a car driving amongst the people. To develop the system to a prototype that can be used to engage with clients, the system will need to be scaled to be able to process thousands of hours of videos and trained to recognise a large array of different behaviours. Massive Analytic will use their expertise in big data solutions to deliver an innovative, robust and computationally efficient video precognition system that can interface easily with third-party systems.

To develop an intelligent recommendation algorithms and controllable prediction engine for automating the selection of suitable data analytical tools for ad hoc (not predefined) analytical tasks, data sources and structures.

PrecogXM is a cloud software application that connects to all the data sources that are important to determine the current purchasing behaviours of customers and their value. It automatically analyses theses data sources and suggests appropriate responses in real-time.Customer Insight teams in large companies use analytics to understand customer behaviour and to measure propensity to buy with an emphasis on developing tailored ‘Just in time’ sales and marketing campaigns. Making sense of what drives customer behaviour is challenging. There are many tools available to customer insight professionals that require deep knowledge of statistics and specialist training. The large volume of data, its great variety and the demand for ‘right time’ marketing, mass customisation and personalisation means that the current costs for gaining actionable customer insights are very high. PrecogXM’s benefits are that it reduces the need for multiple tools and specialist expertise by automating many of the customer insight activities performed by statisticians, thereby reducing costs. PrecogXM ensures customer insight teams can deliver precise offers to their customers at just the right moment. Real time segmentation and behaviour reporting, complex event and transaction processing analytics combined with customer behaviour tracking, are examples of some of the market requirements being addressed.