Tethered Architectures in Cyber-Physical System Development: The Case of Tesla’s Autopilot System

Cyber-physical systems developed to operate under environmental uncertainty, such as self-driving cars, represent challenges within the new product development process. The amount of data and need for iteration of software in addressing the variety and complexity of use environments can conflict with the realities of hardware development lead-time and manufacture. Furthermore, developing new capabilities requires the ability to learn from product use. Current theories regarding product architecture do not provide an adequate architectural blueprint for the development of cyber-physical systems. In this research, through a longitudinal analysis of Tesla's Autopilot system, we propose the concept of a “tethered architecture”. A tethered architecture simultaneously allows the slower, incremental iteration of hardware to be linked with faster, exploratory development of software and continual data-rich feedback loops between product development and use. This paper demonstrates how a tethered product architecture and related development process can enable continuous cycles of learning and adaptation to the variety and complexity of physical environments and situations. This architectural reconfiguration decouples hardware and software while introducing a tight integration between the consumer-facing product and backend software infrastructure and operations. The iterative nature of product development may prove identity-challenging for firms that have traditionally focused on developing complete products, as a tethered architecture gives rise to new product strategies, lifecycle models, knowledge regimes, and organizational arrangements.