This paper focuses on the challenges of vehicle collisions in wireless network-based vehicle driving systems. The identified issues include: 1) data imbalance and control signal loss/delay due to communication, and 2) scheduling failures. In this study, we propose an relatively critical approach upon signal power and controllability to address these problems. Our method includes statistical observations related to communication bottlenecks and analyzes the relationship between communication performance and collision risk. We also refine the existing theo-retical framework for collision prediction to ensure compatibility with current protocols. To evaluate our proposed approach, we set up an environment with two vehicles operating at a fixed distance from each other and analyzed the correlation between communication strength and real-time deadline criticality in relation to collision risk.

Imprecise computing has emerged as an approach for designing energy-efficient digital systems as it can save computing resources. Hot-patching is one of the techniques used to implement it. Hot-patching can add or delete application functions without restarting the application. In this study, we implemented a hot-patching tool for a Linux application and applied it to imprecise computing. We measured the CPU cycle and instruction counts based on square root calculation applications. The experimental results confirmed that CPU resource utilization reduced by 20% when applying a patch with an error range of approximately 2%. Hence, we confirmed that imprecise computing can be performed through hot-patching.

The electronic control unit (ECU), considered the brain of a vehicle, suffers from a design problem called single point of failure (SPOF), which can induce system malfunctions. This problem can be addressed via redundancy, which increases the reliability of a mission-critical system by allowing multiple ECUs to perform a single function. However, this solution requires additional ECU and maintenance costs incurred by the redundant ECUs. A cost-effective approach for improving safety is to utilize the network connectivity between existing ECUs. In this paper, we propose a method that migrates critical tasks residing in an infeasible ECU to a replaceable ECU by using the network connection between them. Furthermore, to demonstrate the feasibility of the method, we implemented a task migration method on a Lego vehicle composed of three ECUs to prevent sudden unintended acceleration accidents caused by faults in an ECU managing the acceleration task.

Software often requires frequent updates to improve performance and reliability. Typically, a general update process is performed after terminating a program although this is not applicable to applications that require non-disruptive services such as networks and satellites. In order to address this issue, network service providers often provide a technology termed as an in-service software upgrade that performs continual updates without stopping the services. However, it requires additional devices or facilities, and the system structure in this case becomes complicated and additional economic costs are incurred. In this study, the design and implementation of a functional-level hot-patching platform are presented for executable and linkable format binary program, based on an ARM and an Intel processor to add or update functions necessary to provide nonstop services. The proposed platform was validated on devices using both Raspberry Pi and a general x86_64 personal computer. Experiments with an iPerf3 server and a drone simulator on each device demonstrated the effectiveness of the proposed hot-patching platform in achieving non-disruptive services with a negligible latency of less than 10 ms.