Analysis of protocols used for visualization in automotive industry

Abstract

One of the biggest obstacles that autonomous driving is facing is the ability to safely and accurately detect and interpret the environment around the vehicle. In other words, detection algorithms need to work perfectly, and for that, the one who made them needs to make sure they work perfectly. One way of debugging autonomous driving algorithms is to use visualization tools that allow you to visually analyze the behavior of the system in real-time. In this paper, Foxglove Studio was chosen as a visualization tool. This work centers around an application that loads video data, takes a frame from it, sends it to a detection algorithm that returns detections, then converts the frame and detections to suitable serialization format and sends it to Foxglove Studio via WebSocket connection, so it can visually display that data to the user. The goal of this work is to determine what is the most suitable mechanism for serializing data in Foxglove Studio so it can be integrated as a part of a bigger platform to help developers. In order to do that, the implementation of 2 different serialization mechanisms was done, and they were compared to each other. Through our testing, we observed that compressing the frame helped to resolve certain issues. As such, we also conducted a performance comparison with and without frame compression. After the analysis was done, it was determined that picking the most suitable format will depend on the specific use case, and that both formats have potential to be used for different reasons. The novelty of this work is that it compares serialization formats in relatively new visualization platform, Foxglove Studio (alpha version was released in June 2021). Even though it is new, it is considered as one of the best tools in robotics and automotive communities because of its powerful visualization and analysis capabilities.