Human drivers navigate the roadways by balancing values such as safety, legality, and mobility. An autonomous vehicle driving on the same roadways as humans likely needs to navigate based on similar values. For engineers of autonomous vehicle technology, the challenge is then to connect these human values to the algorithm design.
To address this challenge, a mapping of philosophical frameworks to mathematical frameworks is used in order to motivate various design choices in a motion planning algorithm. Deontological ethics parallels rule-based mathematical concepts while consequentialism parallels cost-based mathematical concepts. The philosophical theory of virtue ethics is also used to help motivate the relative weightings between the design objectives of path tracking, obstacle avoidance, and adherence to traffic laws. Experimental results of an autonomous vehicle navigating an obstructed two-lane roadway with a double yellow line demonstrate the implications of the various design choices in a model predictive steering controller.
In order to determine the success of the human values captured in an algorithm, the iterative methodology of value sensitive design (VSD) is used to formalize the connection of human values to engineering specifications. A modified VSD methodology is used to develop an autonomous vehicle speed control algorithm to safely navigate a pedestrian crosswalk. Two VSD iterations are presented that model the problem as a partially observable Markov decision process and use dynamic programming to compute an optimal policy to control the longitudinal acceleration of the vehicle based on the belief of a pedestrian crossing. The speed control algorithms are also tested in real-time on an experimental vehicle on a closed-road course.