The rise of Artificial Intelligence (AI) has delivered a huge range of new products aimed at making our lives easier: from voice controlled devices, through fruit picking robots, to autonomous vehicles – these applications bring a wide range of challenges in both development as well as in testing. In this paper, we will consider how we can efficiently test autonomous vehicles to ensure they are safe for the environments where they are required to operate.

The presentation will first consider the various types of AI techniques, such as Bayesian networks; Generic Algorithms; Markov Models; Data Mining; Inductive Logic Programming; and Neural Networks. They represent a wide range of development techniques which have diverse training and testing processes. These will be discussed in terms of how these differ from our traditional software development and testing processes, and thus how our software testing processes need to change.

The focus of the presentation will be on autonomous vehicles. The presenter has been involved in two major projects researching how to test such vehicles to demonstrate they are safe for the context in which they will be used. The first project considers an autonomous POD aimed at last mile journeys (e.g. from car parks to airports or shopping centres) and the second considers a delivery van: how is can drive autonomously around the depot as well as out on the open road. For both projects, we describe how simulated .