Adaptive brake system software platform for self-driving cars

The brake system of a self-driving car is one of the most important systems for ensuring safety. A typical brake system performs several computational tasks, including perception, high-level brake control, and low-level electromechanical control tasks. The status-quo design for scheduling such brake-related tasks is based on the static approach where all parameters for those tasks are fixed when designing the brake system. Such a static approach has the following limitations in terms safety and resource efficiency: i) It cannot adap-tively respond to dynamic environments, such as varying road friction coefficients and the time to collision. ii) The brake operation time constitutes only a small portion of total driving time. Hence, to address this issue, we propose a new adaptive brake system software platform that enables adaptive parameter assignment and dynamic online scheduling to cope with dynamic environments. We implemented and integrated the proposed adaptive parameter assignment and scheduling platform into an AUTOSAR-based brake system, an open and standardized automotive software architecture. Thus, we could significantly improve safety and reliability by shortening the braking distance.