Search Results

Behavioral models of traffic actors have a potential of unlocking sophisticated safety features and mitigating several challenges of urban automated driving. Intuitively, volunteers driving on routes of daily commuting in their private vehicles are the preferred source of information to be captured by data collection system. Such dataset can then serve as a basis for identifying efficient methods of context representation and parameterization of behavioral models. This paper describes the experimental setup supporting the development of driver behavioral models within the SIMUSAFE project. In particular, the paper presents an IoT data acquisition and analysis infrastructure supporting self-confrontation interviews with drivers. The proposed retrofit system was installed in private vehicles of volunteers in two European cities. Wherever possible, the setup used open source software and electronic components available on consumer market.

In the presented paper we deal with an important problem in active safety systems, which is the multi-rate processing of different signals. Automotive systems are usually very complex, involving multiple subsystems, in which typically it is very difficult to obtain equal sampling rates. In many cases, this problem is ignored, which means that the signals samples stored in different time moments are silently assumed to be to sampled in the same time. Looking from the point of view of signal processing, this incorrect assumption often causes large harmonic distortions artifacts of processed signals. These distortions, in turn, generate harmonics of different frequencies. As a result, if processed signals are used to calculate the trajectories of objects seen by systems associated with the vehicle, may differ from the real world trajectories. This may cause occurrence of false positives or no reaction of the vehicle in case of emergency situation.

In this paper we present an example design process of filters used in automotive industry. Signal preprocessing is very important operation in active safety algorithms. Such algorithms usually take into account the vehicle state that includes position, velocities and accelerations of the car. On the basis of these data, as well as the parameters and trajectories of external objects “observed” by the car, the algorithms make decisions about various safety actions. Designer of such algorithms must assure an appropriate quality of such signals, which usually means a proper filtering. In this paper we focus on selected important aspects of the filter design process. The main objectives of the presented investigations is to obtain such filters that ensure a sufficient rejection of undesired components from the signal and at the same time that do not introduce too high delay to the processed signals.

Active Safety (AS) and Advanced Driver Assistance Systems (ADAS) can nowadays be considered as distributed embedded software systems where independent microprocessor systems communicate together using different communication protocols. Typical AS or ADAS functionality is then realized by several microprocessors communicating with each other. AS and ADAS systems interact with other Electronic Control Units in a vehicle via communication networks and gather vehicle's surroundings via camera, radar or laser sensors. Quality assurance and safety standards combined with increasing complexity and reliability demands related to vision sensing, radar sensing and data fusion, often together with a short time to market, make the development of such systems challenging. As the number of important road scenarios for the system grows, mathematical modelling and computer simulation become important engineering tasks that aim to assure the required quality and compliance with safety standards.