Page 134 - Dynamic Vision for Perception and Control of Motion
P. 134

118       4  Application Domains, Missions, and Situations


            without using indication signs (blinker); even these critical situations not conform-
            ing to standard behavior have to be coped with successfully.


            4.3  Situations as Precise Decision Scenarios


            The definition for “situation” used here is the following: A situation encompasses
            all aspects of relevance for decision-making in a given scenario and mission con-
            text. This includes environmental conditions affecting perception and limit values
            for control application (such as wheel to ground friction coefficients) as well as the
            set of traffic regulations actually valid that have been announced by traffic signs
            (maximum speed allowed,  passing  prohibited,  etc.). With  respect  to other ob-
            jects/subjects, a situation is not characterized by a single relation to one other unit
            but to the  total number of  objects  of relevance.  Which  of those detected and
            tracked are relevant is a difficult decision. Even the selected regions of special at-
            tention are of importance. The objects/subjects of relevance are not necessarily the
            nearest ones; for example, driving at higher speed, some event happening at a far-
            ther look-ahead distance than the two preceding vehicles may be of importance: A
            patch of dense fog or a front of heavy rain or snow can be detected reliably at rela-
            tively long distance. One should start reacting to these signs at a safe distance ac-
            cording to independent judgment and not only when the preceding vehicles start
            their reactions.
              Some situational aspects can be taken into account during mission planning. For
            example, driving on roads heading into the low-standing sun at morning or evening
            should  be avoided  by  proper selection  of  travel time. Traffic congestion  during
            rush hour also may be avoided by proper timing. Otherwise, the driver/autonomous
            vehicle has to perceive the indicators for situational aspects, and from a knowledge
            base, the proper behavior has to be selected. The three components required to per-
            form this reliably are discussed in the sections below: Environmental background,
            objects/subjects of relevance, and the rule systems for decision-making. Beside the
            rules for handling planned  missions, another set of perceptual  events has to be
            monitored which may require another set of rules to be handled for selecting
            proper reactions to these events.



            4.3.1  Environmental Background
            This has not received sufficient attention in the recent past since, at first, the basic
            capabilities of perceiving roads and lanes as well as other vehicles had to be dem-
            onstrated. Computing power for including at least some basic aspects of environ-
            mental conditions at reasonable costs is now coming along. In Section 4.1 and Ap-
            pendix A.1.2 (lighting conditions)/A.1.3 (weather conditions), some aspects have
            already been mentioned. Since these environmental conditions change rather
            slowly, they may be perceived at a low rate (in the range of seconds to minutes).
            An economical way to achieve this may be to allot remaining processing time per
            video cycle of otherwise dedicated image processing computers to this “environ-
   129   130   131   132   133   134   135   136   137   138   139