Safety assessment of automated vehicles: how to determine whether we have collected enough field data?

Abstract

Objective: The amount of collected field data from naturalistic driving studies is quickly increasing. The data are used for, among others, developing automated driving technologies (such as crash avoidance systems), studying driver interaction with such technologies, and gaining insights into the variety of scenarios in real-world traffic. Because data collection is time consuming and requires high investments and resources, questions like “Do we have enough data?,” “How much more information can we gain when obtaining more data?,” and “How far are we from obtaining completeness?” are highly relevant. In fact, deducing safety claims based on collected data—for example, through testing scenarios based on collected data—requires knowledge about the degree of completeness of the data used. We propose a method for quantifying the completeness of the so-called activities in a data set. This enables us to partly answer the aforementioned questions.

Method: In this article, the (traffic) data are interpreted as a sequence of different so-called scenarios that can be grouped into a finite set of scenario classes. The building blocks of scenarios are the activities. For every activity, there exists a parameterization that encodes all information in the data of each recorded activity. For each type of activity, we estimate a probability density function (pdf) of the associated parameters. Our proposed method quantifies the degree of completeness of a data set using the estimated pdfs.

Results: To illustrate the proposed method, 2 different case studies are presented. First, a case study with an artificial data set, of which the underlying pdfs are known, is carried out to illustrate that the proposed method correctly quantifies the completeness of the activities. Next, a case study with real-world data is performed to quantify the degree of completeness of the acquired data for which the true pdfs are unknown.

Conclusion: The presented case studies illustrate that the proposed method is able to quantify the degree of completeness of a small set of field data and can be used to deduce whether sufficient data have been collected for the purpose of the field study. Future work will focus on applying the proposed method to larger data sets. The proposed method will be used to evaluate the level of completeness of the data collection on Singaporean roads, aimed at defining relevant test cases for the autonomous vehicle road approval procedure that is being developed in Singapore.     

Resource protection for waterbirds in Chesapeake bay

Many living resources in the Chesapeake Bay estuary have deteriorated over the past 50 years. As a result, many governmental committees, task forces, and management plans have been established. Most of the recommendations for implementing a bay cleanup focus on reducing sediments and nutrient flow into the watershed. We emphasize that habitat requirements other than water quality are necessary for the recovery of much of the bay's avian wildlife, and we use a waterbird example as illustration. Some of these needs are: (1) protection of fast-eroding islands, or creation of new ones by dredge deposition to improve nesting habitat for American black ducks(Anas rubripes), great blue herons(Ardea herodias), and other associated wading birds; (2) conservation of remaining brackish marshes, especially near riparian areas, for feeding black ducks, wading birds, and wood ducks(Aix sponsa); (3) establishment of sanctuaries in open-water, littoral zones to protect feeding and/or roosting areas for diving ducks such as canvasbacks(Aythya valisineria) and redheads(Aythya americana), and for bald eagles(Haliaeetus leucocephalus); and (4) limitation of disturbance by boaters around nesting islands and open-water feeding areas. Land (or water) protection measures for waterbirds need to include units at several different spatial scales, ranging from “points” (e.g., a colony site) to large-area resources (e.g., a marsh or tributary for feeding). Planning to conserve large areas of both land and water can be achieved following a biosphere reserve model. Existing interagency committees in the Chesapeake Bay Program could be more effective in developing such a model for wildlife and fisheries resources.     

A novel cell division factor from tobacco 2B-13 cells that induced cell division in auxin-starved tobacco BY-2 cells

Effects of auxin as plant hormones are widespread; in fact in almost all aspects of plant growth and development auxin plays a pivotal role. Although auxin is required for propagating cell division in plant cells, its effect upon cell division is least understood. If auxin is depleted from the culture medium, cultured cells cease to divide. It has been demonstrated in this context that the addition of auxin to auxin-starved nondividing tobacco BY-2 cells induced semisynchronous cell division. On the other hand, there are some cell lines, named habituated cells, that can grow without auxin. The cause and reason for the habituated cells have not been clarified. A habituated cell line named 2B-13 is derived from the tobacco BY-2 cell line, which has been most intensively studied among plant cell lines. When we tried to find the difference between two cell lines of BY-2 and 2B-13 cells, we found that the addition of culture filtrated from the auxin-habituated 2B-13 cells induced semisynchronous cell division in auxin-starved BY-2 cells. The cell division factor (CDF) that is responsible for inducing cell division in auxin-starved BY-2 cells was purified to near-homogeneity by sequential passage through a hydroxyapatite column, a ConA Sepharose column and a Sephadex gel filtration column. The resulting purified fraction appeared as a single band of high molecular weight on sodium dodecyl sulfate-polyacrylamide gel electrophoresis gels by silver staining and was able to induce cell division in auxin-starved BY-2 cells. Identification of the protein by MALD-TOF-MS/MS revealed that it is structurally related to P-glycoprotein from Gossypioides kirkii, which belongs to ATP-binding cassette (ABC)-transporters. The significance of CDF as a possible ABC-transporter is discussed in relationship to auxin–autotrophic growth and auxin-signaling pathway.     

Response to commentary on “The Driver Behaviour Questionnaire as a predictor of accidents: A meta-analysis”

A commentary on our previously published meta-analysis about the predictive validity of the Driver Behaviour Questionnaire (DBQ) raised a number of points. These points do not dispute the quantitative results as such, but suggest that our introduction and discussion overly favor the DBQ and are incomplete in a number of ways. The commentary targeted the following topics: common method variance, intercorrelations of different instruments, accident data validity, correcting for measurement error,

Common method variance

Our meta-analysis provided an extensive discussion of validity threats, including ones not treated before in the DBQ literature, such as common scale anchors and publication bias. We see little added value in the commentary when it informs the readership of common method variance (CMV), as we clearly did this in our article. It is widely understood that CMV can account for a large share of the variance when self-reported data are intercorrelated (see Podsakoff, MacKenzie, Lee, & Podsakoff, 2003

Intercorrelations of different instruments

The commentary reacts to an introductory sentence in which we stated that the DBQ is strongly situated in a network of other questionnaires and tests (such as Trait Anxiety, Cognitive Failures Questionnaire, and Sensation Seeking Scale), by asserting that we seem to interpret such correlations as a positive feature of the DBQ and by pointing out that correlations between the DBQ and other self-reports may have arisen spuriously because of CMV.We dispute the assertion that we regard correlations

Accident data validity

The commentary rejects our position that not only self-reported accident data are susceptible to biases, but recorded data too. Note that our remark applied to all types of recorded accidents, including police reports, hospital data, insurance data, as well as fleet data from professional drivers, and not just company data as in the work by af Wåhlberg, Dorn, and Kline (2011), which is mentioned in the commentary. The literature discusses several sources of bias for recorded accidents that are

Correcting for unreliability

The commentary pointed out that our correction for attenuation is unusual, a surprising claim considering the established importance of correction for measurement error in theory testing (Liu and Salvendy, 2009, Schmidt and Hunter, 1999). The available DBQ research provided almost no information on measurement error, so we did not apply a correction as part of the meta-analysis. Instead, we applied one afterwards, based on the raw data of the largest DBQ study available, and illustrated that

Exposure

The commentary points out that our results were not corrected for exposure in any way. First of all, this statement is false, because our meta-analysis did include effect sizes corrected for exposure. We used a special moderator category for effect sizes other than zero-order correlations. These effect sizes were derived from regression analysis, often with mileage as one of the predictors.Second, it may be noted that af Wåhlberg himself reported that the association between exposure and

Further references

The commentary attended us to 11 studies supposedly not included in our meta-analysis. We appreciate af Wåhlberg and Dorn's close scrutiny of our reference list for potential omissions on the DBQ-accident relationship. It may be noted that missing a small fraction of studies in a meta-analysis of this scope is almost inevitable.The samples of four of the studies (Dobson et al., 1999, Elliott et al., 2007, Parker, 1999, Stradling et al., 2005) identified in the commentary were already included

Discussion

The purpose of a meta-analysis is to provide a quantitative summary of a metric of interest, correlations between the DBQ, and external criteria in our case. af Wåhlberg and Dorn's commentary does not dispute the quantitative results in themselves (except with regard to the correction for exposure), but targets the qualitative introduction and discussion of our article.The commentary raises some valid points, albeit points already discussed in our article. We too believe that common method

     

Organoclays for Aquifer Bioremediation: Adsorption of Chlorobenzene on Organoclays and its Degradation by RHODOCOCCUS B528

The adsorption and degradation of chlorobenzene on partially modified organoclays and by the autochthonous microorganism Rhodococcus B528 were studied by means of the batch technique. Organoclays were prepared from Na-montmorillonite (MM) by using dodecyltrimethylammonium (C₁₂) and dioctadecyldimethylammonium (2C₁₈) bromides. The degree of modification was 35 (2C₁₈-35-MM) and 89% (C₁₂-89-MM) of the cation exchange capacity of MM. The adsorption experiments were carried out using headspace GC. The intercalation of chlorobenzene into the interlayers of organo-MM was detected by X-ray diffraction. The adsorption isotherms found were of the S1 type indicating a cooperative effect. Chlorobenzene showed a higher affinity for 2C₁₈-35-MM than C₁₂-89-MM, which could not only be explained by the organic carbon content. The comparison with 2,4-dichlorophenol adsorption has implied that for the studied systems the different adsorption mechanisms are primarily governed by the different molecular properties and not by the type of absorbent. The presence of 2C₁₈-35-MM caused no negative effect on the investigated microorganisms and complete biodegradation of chlorobenzene was achieved without desorption limitation for growth, demonstrating the applicability of partially modified organoclays for bioremediation.     

Über den Einfluß der Nahrungskonzentration und anderer Faktoren auf Filtrierleistung und Nahrungsausnutzung der Muscheln Arctica islandica und Modiolus modiolus

Filtration rates and the extent of phagocytosed food particles were determined in the offshore lamellibranchs Artica islandica and Modiolus modiolus in relation to particle concentration, body size and temperature. Pure cultures of the algae Chlamydomonas sp. and Dunaliella sp. were used as food. A new method for determining filtration rates was developed by modifying the classical indirect method. The concentration of the experimental medium (100%) was kept constant to ±1%. Whenever the bivalves removed algae from the medium, additional algae were added and the filtration rate of the bivalves expressed in terms of percentage amount of algae added per unit time. The concentration of the experimental medium was measured continuously by a flow colorimeter. By keeping the concentration constant, filtration rates could be determined even in relation to different definite concentrations and over long periods of time. The amount of phagocytosed food was measured by employing the biuret-method (algae cells ingested minus algae cells in faeces). Filtration rates vary continuously. As a rule, however, during a period of 24 h, two phases of high food consumption alternate with two phases of low food consumption during which the mussels' activities are almost exclusively occupied by food digestion. Filtration rate and amount of phagocytosed algae increase with increasing body size. Specimens of A. islandica with a body length of 33 to 83 mm filter between 0.7 to 71/h (30–280 mg dry weight of algae/24 h) and phagocytose 21 to 122 mg dry weight of algae during a period of 24 h. The extent of food utilization declines from 75 to 43% with increasing body size. In M. modiolus of 40 to 88 mm body length, the corresponding values of filtration rate and amount of phagocytosed algae range between 0.5 and 2.5 l/h (20–100 mg dry weight of algae) and 17 to 90 mg dry weight of algae, respectively; the percentage of food utilization does not vary much and lies near 87%. Filtration rate and amount of phagocytosed algae follow the allometric equation y=a·x ^b. In this equation, y represents the filtration rate (or the amount of phagocytosed algae), a the specific capacity of a mussel of 1 g soft parts (wet weight), x the wet weight of the bivalves' soft parts, and b the specific form of relationship between body size and filtration rate (or the amount of phagocytosed algae). The values obtained for b lie within a range which indicates that the filtration rate (or the amount of phagocytosed algae) is sometimes more or less proportional to body surface area, sometimes to body weight. Temperature coefficients for the filtration rate are in Arctica islandica Q₁₀ (4°–14°C)=2.05 and Q₁₀ (10°–20°C)=1.23, in Modiolus modiolus Q₁₀ (4°–14°C)=2.33 and Q₁₀ (10°–20°C)=1.63. In A. islandica, temperature coefficients for the amount of phagocytosed algae amount to Q₁₀ (4°–14°C)=2.15 and Q₁₀ (10°–20°C)=1.55, in M. modiolus to Q₁₀ (4°–14°C)=2.54 and Q₁₀ (10°–20°C)=1.92. Upon a temperature decrease from 12° to 4°C, filtration rate and amount of phagocytosed algae are reduced to 50%. At the increasing concentrations of 10×10⁶, 20×10⁶ and 40×10⁶ cells of Chlamydomonas/l offered, filtration rates of both mollusc species decrease at the ratios 3:2:1. At 12°C, pseudofaeces production occurs in both species in a suspension of 40×10⁶, at 20°C in 60×10⁶ cells of Chlamydomonas/l. At 12°C and 10–20×10⁶ cells of Chlamydomonas/l, the maximum amount of algae is phagocytosed. At 40×10⁶ cells/l, the amount of phagocytosed cells is reduced by 26% as a consequence of low filtration rates and intensive production of pseudofaeces. At 20°C and 20–50×10⁶ cells of Chlamydomonas/l, the maximum amount of algae is sieved out and phagocytosed; the concentration of 10×10⁶ cells/l is too low and cannot be compensated for by increased activity of the molluscs. With increasing temperatures, the amount of suspended matter, allowing higher rates of filtration and food utilization, shifts toward higher particle concentrations; but at each temperature a threshold exists, above which increase in particle density is not followed by increase in the amount of particles ingested. Based on theoretical considerations and facts known from literature, 7 different levels of food concentration are distinguishable. Experiments with Chlamydomonas sp. and Dunaliella sp. used as food, reveal the combined influence of particle concentration and particle size on filtration rate. Supplementary experiments with Mytilus edulis resulted in filtration rates similar to those obtained for M. modiolus, whereas, experiments with Cardium edule, Mya arenaria, Mya truncata and Venerupis pullastra revealed low filtration rates. These species, inhabiting waters with high seston contents, seem to be adapted to higher food concentrations, and unable to compensate for low concentrations by higher filtration activities. Adaptation to higher food concentrations makes it possible to ingest large amounts of particles even at low filtration rates. Suspension feeding bivalves are subdivided into four groups on the basis of their different food filtration behaviour.