ITS and Road Safety

       Safety occupies a central place in the scopes of Intelligent Transportation Systems. Unlike other function types, such as comfort, routing, traffic management, etc., safety may involve critical interactions with the driver, touching the issue of who retains the authority. This is especially true for preventive safety applications addressing the tactical time frame, where control of the vehicle by the intelligent system greatly overlaps driver capabilities.
       The following selection of papers, far from being a complete picture of al safety relevant aspects in ITS, has been chosen to give a peek of recent advances and, above all, to outline future trends and possible roadmap for research.
        The leitmotif is the coexistence of intelligent system and man, which develops in several intertwined threads.
      The first thread is perception. Accurate and reliable perception is a prerequisite for safety-critical applications. Perception has been the first, and probably most researched, topic and most has been written. To summarize, paper [1] has been chosen, which is an editorial of a special journal issue on perception navigation and planning, making a good introduction of the role of perception for intelligent transportation systems.
        A second thread is the variety of safety support functions, which is embodied in many recent new generation prototypal ADAS. Papers [2]-[5] are from the European project PReVENT and give an idea of the large spectrum and complexity of safety support functions, such as early warning [2], longitudinal and lateral maneuvering [3], [4] and holistic driver support [5].
        A third thread concerns the presentation of information to the driver, human factors and human centric design of interactions. Two papers have been selected to present this perspective: one is from the European Project AIDE [6], another from US (NSF) [7].
        The above three categories can be viewed as corresponding to a “sense-think-act” paradigm (sense the world, assess the situation, warn or intervene), which is the most immediate and natural scheme for safety driver assistance system. However, further perspectives are emerging. Let us see these ideas, projecting into the future.
        A fourth thread is related to the collaboration of a man and a system and later between man-systems, which brings about a number of other facets.
        One of these is the idea that the system may play the role of a co-pilot. Strictly speaking a copilot would be the second pilot of an aircraft, which means that the intelligent system is intended to relieve some driving tasks, with a variable degree of automation. The project HAVEit investigated a joint system formed with such a co-pilot [8], [9] and in particular the dynamic (re)allocation of tasks between driver and co-pilot and the transitions between levels of automation.
        A variation of the co-pilot idea may be found in [3], where the system embeds a “virtual user”. Instead of being focused on co-driving, virtual users put the emphasis on modeling how human driver would drive, producing humanlike “reference maneuvers” as a sound basis to design support functions. Virtual users can be extended to the analysis of alternative maneuvers to provide integrated holistic support [5].
        A fifth thread is related to another facet of the man-system interaction, which is close to the idea of modeling driver as a means to support them. Namely: prediction of driver “intentions”. The idea is to recognize preparatory motor cues that anticipate some type of manure, and thus to support the driver for the proper maneuver and with sufficient anticipation. A survey of these methods can be found in [10].
        A sixth embryonic thread could be referred as “understanding” the driver which means a further step where driver models (e.g., virtual users) and intention predictions are fused, so that maneuvers are explained and linked to driver goals (i.e., not only what is imminent but also why). Understanding driver goals might become the “engine” of both effective joint systems and of effective cooperative systems too, where agents “know” the goals of each other. Peeks on these trends are given by paper [11]-[14]. Paper [11] describes the advantages of sharing “intentions” in cooperative systems and also explicitly mentions human centric trajectory prediction (i.e., a human user model). Paper [12] presents the idea of “cognitive vehicles”, which are intended as vehicles designed on driver cognition models. Paper [13] presents an application of an artificial cognitive system learning driver models (i.e., forming an internal virtual user model) and using such models as a basis for interactions, namely: to warn the driver when departing excessively from the learnt model. Lastly paper [14] presents path predictions that are based on expected driver input, i.e., on an internal driver model.
        We wish you an enjoyable reading of this special issue.

The Editors: Dr. Angelos Amditis, Dr. Mauro Da Lio and Mr. Roy Goudy

Selected Articles

Leave a Reply