The Future of Legal and Ethical Regulations for Autonomous Robotics Huan Xu 1 and Joseph E. Borson 2 Abstract— “Autonomous robotics” promise significant im- provements across a host of different complex systems, which will need to be managed within regulatory frameworks to promote, at a minimum, device safety. Contrary to how they are often portrayed, however, these systems do not necessarily require fundamentally new approaches to engineering or reg- ulatory challenges, i.e., the development of a novel “autonomy framework” applicable to different types of devices. Rather, because autonomous systems generally represent a progressive improvement of existing complex systems, preexisting regu- latory scheme offer the best guidance for considering future regulation of autonomous elements. Moreover, the regulatory landscape differs considerably based on the type of device at issue (e.g., consumer electronics vis-` a-vis medical devices). This paper argues that users and regulators must consider future autonomy regulations within the specific framework those devices currently inhabit, rather than focusing on a novel set of rules divorced from the preexisting context. I. INTRODUCTION “Autonomous robotics,” at least as popularly construed in the vein of self-driving cars, swarms of flying drones, or even military UAVs capable of deciding on their own to fire missiles, has been often been thought to be a fundamentally new type of technology, requiring fundamentally new types of engineering principles and regulatory approaches - an “autonomy” approach, as it were. This thought piece argues the opposite - autonomy is merely an evolutionary outgrowth of existing systems, and so existing systems engineering and regulatory approaches are the best ways of governing them. It does not make sense to speak of “autonomy” law, regulation, or even systems design as a new-in-kind category, or one that is generally applicable to different domains (e.g., aviation, automobiles, consumer products, etc.). Rather, ex- isting frameworks - in particular, existing means of regulating complex systems (with their commensurate incorporation of different values and risk tolerances) are the best means to regulate autonomous systems. There is not a universal definition of an autonomous system. Autonomy, however, at its core is merely the process of designing or regulating systems that make choices based not on contemporaneous, explicit, human intervention, but on pre-conceived decision rules. A self-driving car that de- termines how best to navigate a route based on an algorithm and sensor inputs is autonomous, but so is, in some sense, a stoplight that changes color based on pressure sensors, or even a medical device that delivers medicine based on a series of biological inputs. Accordingly, ”autonomy” is not 1 Huan Xu is with Faculty of Aerospace Engineering, University of Maryland, College Park, MD, 20742, USA. mumu@umd.edu 2 Joseph Borson is a graduate of Harvard Law School, Washington, DC, 20009, USA. jborson@jd13.law.harvard.edu something new, rather, it is a measure of complexity that the law and the engineering fields have been confronting for many years. Importantly, different fields have addressed autonomy challenges in different ways. Some fields, such as aviation, may prohibit autonomous devices altogether absent specific approval. Other fields, such as consumer electronics, may al- low autonomous devices so long as they do not affirmatively interfere or cause problems. These approaches are often governed by the relative risk and benefits of the systems, or in some cases, by regulatory inertia or path dependency. Fields, for example, that are more heavily regulated or that require regulator pre-approval to begin with, like for certain medical devices, will follow these structural requirements with autonomous systems, just as they would with any system. Key, however, is that there is no single approach to regulation of autonomous systems - it is system and regulator specific. Accordingly, attempts to develop a unified theory of autonomous regulation, to the extent that such attempts are ongoing, are bound to fail, for the simple reason that they are necessarily unbound from the specific context of the systems they are trying to govern. This work investigates what the legal and engineering communities believe are the current challenges in regulating autonomous systems, and posits that those challenges are not the true difficulties in legal and ethical regulation. This paper is structured as follows: Section II discusses the current challenges to regulation, Section III describes three application areas, and Section IV presents our conclusions. II. CURRENT CHALLENGES While there have been a number of articles in the news regarding the oncoming rush of autonomous systems into daily life, most have only posed concerns regarding what those changes could imply [1]. As the issue of regulation in autonomy is fairly new, the preliminary available literature in this area is somewhat limited. However, a few broad categories seem to emerge from existing work. These topics fall under the following categories: ethics, liability, and safety. We present a brief description of each of these issues, as well as reasons why the focus on these particular issues do not specifically address actual regulatory decisions. A. Ethics and Society 1) Previous Work: A number of works focus on the issue of imposing morals on machines [2]. Anderson et al. [3] discusses the moral reasons why autonomous machines should function ethically. In the area of medical robotics, Stahl and Coeckelbergh [4] explore the implications of health 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) Madrid, Spain, October 1-5, 2018 978-1-5386-8094-0/18/$31.00 ©2018 IEEE 2362