Article

Investigating the effects of roadside cover on safe speeds for autonomous driving in high-risk deer-vehicle collision areas

Authors:
To read the full-text of this research, you can request a copy directly from the authors.

Abstract

This paper explores the effects of roadside vegetation and/or cover on the probability of vehicular collisions with deer by identifying and simulating dangerous animal-vehicle interaction scenarios. Deer motion is described by a simplistic model inspired by the literature on deer escape behavior, and a genetic algorithm identifies problematic combinations of model parameter values given a particular vehicle-driver-environment configuration. A nonlinear, planar bicycle model is used to simulate vehicle dynamics. Two simulated driver-assist systems drive the vehicle model in simulation: the first applies brakes to approximate a “forward collision assist” system, and the second combines braking and steering inputs using model predictive control to avoid obstacles. Simulations of these driver-vehicle models interacting with deer are carried out for various configurations of roadside cover geometry. For the configurations explored in this study, the methodology produces recommended safe driving speeds for vehicles employing each driver assist system given a particular road configuration.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the authors.

Preprint
Full-text available
Autonomous vehicles (AV) are expected to play a key role in the future of transportation, introducing a disruptive yet potentially beneficial change for vehicle-wildlife interactions. However, this assumption has not been critically examined. Here, we introduce a new conceptual framework covering the intersection between AV technological innovation and wildlife conservation to reduce wildlife-vehicle collisions. We suggest future research within this framework to focus on developing robust warning systems and animal detection methods for AV systems, and to incorporate wildlife-vehicle interactions into decision-making algorithms. With large-scale deployment a looming reality, it is vital to incorporate conservation and sustainability into the societal, ethical, and legal implications of AV technology. We appeal for further debate and interdisciplinary collaborations between scientists, developers, and policymakers.
Article
Full-text available
Most deer–vehicle collisions (DVCs) occur at night when deer are active and the ability of motorists to see them is impaired. The objective of this study was to examine the ability of motorists to detect deer at night and examine select factors which may influence detection distances. We examined the ability of motorists to detect deer at night and the factors that influence detection using deer decoys, volunteer motorists, and a series of driving courses. Deer decoys were detected at significantly greater distances when located on the right rather than on the left side of the road and when
Article
Full-text available
Survival and cause-specific mortality of female white-tailed deer (Odocoileus virginianus) have been well documented in forested and agricultural landscapes, but limited information has been collected in grassland habitats typical of the Northern Great Plains. Our objectives were to document and compare survival and cause-specific mortality of adult female white-tailed deer in four distinct ecoregions. We captured and radiocollared 190 (159 adult, 31 yearling) female white-tailed deer and monitored (including deer from a previous study) a total of 246 (215 adult, 31 yearling) deer from Jan. 2000 to Dec. 2007. We documented 113 mortalities; hunting (including wounding loss) accounted for 69.9% of all mortalities and vehicle collisions accounted for an additional 15.0%. Natural causes (e.g., disease, predation) of mortality were minor compared to human-related causes (e.g., hunting, vehicle collisions). We used known fate modeling in program MARK to estimate survival rates and compare ecoregions and seasons. Model {Sseason (winter = summer)} had the lowest AICc value suggesting that survival differed only between seasons where winter and summer survival was equal and differed with fall season. Annual and seasonal (summer, fall, winter) survival rates using the top model {Sseason (summer = winter)} were 0.76 (95% ci  =  0.70–0.80), 0.97 (95% ci  =  0.96–0.98), 0.80 (95% ci  =  0.76–0.83) and 0.97 (95% ci  =  0.96–0.98), respectively. High human-related mortality was likely associated with limited permanent cover, extensive road networks and high hunter density. Deer management in four distinct ecoregions relies on hunter harvest to maintain deer populations within state management goals.
Article
Full-text available
Ungulates are viewed as being highly susceptible to predation during the initial weeks or months of life. Yet aggressive defence by adult females is common in many ungulates and has the potential to reduce the vulnerability of the young significantly. We observed naturally occurring predatory encounters between coyotes Canis latrans Say and deer fawns to test the hypothesis that a difference in aggressive defence leads to the differential vulnerability of mule deer and white-tailed deer fawns in summer, when fawns are 0-14 weeks in age. Whitetail fawns suffer higher levels of coyote predation than do mule deer fawns at that time. The two species of deer are similar in size, but are known to differ in their antipredator behaviour in winter when fawns are older. Coyotes were less likely to attack mule deer than whitetail fawns they encountered, and were less likely to kill mule deer than whitetail fawns they attacked. The presence of a mule deer, but not a whitetail, female with a fawn deterred coyotes f
Article
Full-text available
Collisions between deer and vehicles have become very common and pose a safety problem to deer as well as to motorists (Désiré & Recorbet 1985, Kofler. 2004: Effectiveness of acoustic road markings in reducing deer-vehicle collisions: a behavioural study. -Wildl. Biol. 10: 155-159. The behavioural responses and habituation of fallow deer Dama dama to play-back sounds from repeatedly occurring acoustic road markings were stud-ied during 13 nights. Our experimental design eliminated factors normally asso-ciated with passing vehicles (e.g. vehicle noise and light), and fallow deer were exposed to play-back sounds from road markings at predetermined time inter-vals. Though the distribution of the predefined behavioural responses 'flight', 'alarm', 'movement of head' and 'no reaction' varied among nights, the fallow deer exhibited increasing indifference to sounds from road markings with time, and this we explained as being habituation to the acoustic stimulus. As we expect our results to be valid for other species of deer as well, we find that acoustic road markings are not a reliable method to reduce the number of deer-vehicle collisions on a long-term basis.
Article
Full-text available
In predator-prey encounters, many factors influence risk perception by prey and their decision to flee. Previous studies indicate that prey take flight at longer distances when they detect predators at longer distances and when the predator's behavior indicates the increased likelihood of attack. We examined the flight decisions of Columbian black-tailed deer (Odocoileus hemionus columbianus) using an approaching human whose speed, directness of approach, directness of gaze, and simulated gun carrying varied. Deer fled at greater distances when approached more quickly and directly, and there was a concave-down quadratic trend in the relationship between the distances at which the predator began its approach and at which the deer became alert (alert distance [AD]), indicating that deer have a zone of awareness beyond which there is a delay in detecting an approaching predator. Time spent assessing the approacher (assessment time) was shorter during faster approaches and was positively related with AD. Deer fled at longer distances and had shorter assessment times when they were already alert to the predator at the initiation of approach. Males fled at shorter distances than females when approached during the gun-holding condition, and males had shorter assessment times than females when the approacher averted his gaze. Such sex differences in risk assessment might reflect male motivation during the mating season as well as exposure to human hunting. We suggest that risk assessment is affected the by the predator's behavior, the state of awareness of the prey, and the distance at which they detect the predator. Copyright 2006.
Article
Human-driven and autonomously driven cars of today act often reactively to the decisions of the cars they follow, which could lead to uncomfortable, inefficient, and sometimes unsafe situations in stop and go traffic. This paper proposes methods for probabilistic anticipation of the motion of the preceding vehicle and for the control of motion of the ego vehicle. We construct: 1) a Markov chain predictor based on the observed behavior of preceding vehicle and 2) a maximum likelihood motion predictor based on historical traffic speed at different locations and times. Heuristics are proposed for combining the two predictions to determine a probability distribution on the position of the preceding vehicle over a future planning horizon. A chance-constrained model predictive control framework is employed to optimize the motion of the ego vehicle, given the probabilistic prediction of motion of preceding vehicle. Effectiveness of the proposed approach is evaluated in multiple simulation scenarios.
Book
The definitive book on tire mechanics by the acknowledged world expert. © 2012 Hans Pacejka Published by Elsevier Ltd All rights reserved.
Article
Self-driving vehicles are a maturing technology with the potential to reshape mobility by enhancing the safety, accessibility, efficiency, and convenience of automotive transportation. Safety-critical tasks that must be executed by a self-driving vehicle include planning of motions through a dynamic environment shared with other vehicles and pedestrians, and their robust executions via feedback control. The objective of this paper is to survey the current state of the art on planning and control algorithms with particular regard to the urban setting. A selection of proposed techniques is reviewed along with a discussion of their effectiveness. The surveyed approaches differ in the vehicle mobility model used, in assumptions on the structure of the environment, and in computational requirements. The side-by-side comparison presented in this survey helps to gain insight into the strengths and limitations of the reviewed approaches and assists with system level design choices.
Article
During 1986-1988, mother-young relations in white-tailed deer (Odocoileus virginianus) were studied at the National Zoological Park's Conservation and Research Center. During the first few weeks after parturition, mothers associated very little with their fawns and periods of contact were brief; they visited their fawns predominantly at sunrise and sunset. Later, fawns were visited randomly, and time of contacts and length of time between successive contacts became variable. Mothers kept the shortest distances during the first 2 weeks; distances then increased and remained constant. Up to 4 weeks of age, distances were shortest at night, but later, mothers were located farthest from their fawns during this time. Mother-young distances during times of reunion also increased with age. Behavioral differences between sexes began to unfold in mother-young relationships as early as 3 weeks after birth and persisted throughout the first 3 months. Males associated less and more loosely with their mothers than females.
Article
During the summers of 1980-82 the extent, causes, and timing of white-tailed deer (Odocoileus virginianus) fawn mortality were investigated on Crab Orchard National Wildlife Refuge (CONWR), southern Illinois. Summer mortality rates for 54 radio-collared fawns averaged 30%. Coyotes (Canis latrans) and domestic dogs accounted for 69% of natural mortalities. Most fawns lost to canids were 27-47 days old.
Article
Nineteen habitat and highway characteristics thought to influence numbers of white-tailed deer (Odocoileus virginianus) vehicle collisions along Pennsylvania two-lane highways were tested for inclusion in a model used to predict probabilities of sections of highways being high deer kill sites. Two variables (in-line visibility and non-wooded) increased the probability of a section of highway being a high kill site. Seven variables (residences, commercial buildings, other buildings, shortest visibility, speed limit, distance to woodland, and fencing) decreased this probability. Removal of the variables speed limit and other buildings did not significantly change the model. The model showed strong discrimination between high kill and low kill ("control") sections of highway.
Conference Paper
In this paper, we present the strategy for trajectory planning that was used on-board the vehicle that completed the 103 km of the Bertha-Benz-Memorial-Route fully autonomously. We suggest a local, continuous method that is derived from a variational formulation. The solution trajectory is the constrained extremum of an objective function that is designed to express dynamic feasibility and comfort. Static and dynamic obstacle constraints are incorporated in the form of polygons. The constraints are carefully designed to ensure that the solution converges to a single, global optimum.
Article
Introduction This paper presents the results of a comprehensive project whose goal is to identify roadway design practices that maximize the margin of safety between the friction supply and friction demand. This study is motivated by the concern for increased accident rates on curves with steep downgrades, geometries that contain features that interact in all three dimensions — planar curves, grade, and superelevation. This complexity makes the prediction of vehicle skidding quite difficult, particularly for simple simulation models that have historically been used for road geometry design guidance. Method To obtain estimates of friction margin, this study considers a range of vehicle models, including: a point-mass model used by the American Association of State Highway Transportation Officials (AASHTO) design policy, a steady-state “bicycle model” formulation that considers only per-axle forces, a transient formulation of the bicycle model commonly used in vehicle stability control systems, and finally, a full multi-body simulation (CarSim and TruckSim) regularly used in the automotive industry for high-fidelity vehicle behavior prediction. The presence of skidding – the friction demand exceeding supply – was calculated for each model considering a wide range of vehicles and road situations. Results The results indicate that the most complicated vehicle models are generally unnecessary for predicting skidding events. However, there are specific maneuvers, namely braking events within lane changes and curves, which consistently predict the worst-case friction margins across all models. This suggests that any vehicle model used for roadway safety analysis should include the effects of combined cornering and braking.
Conference Paper
In rural areas, wildlife animal road crossings are a threat to both the driver and the wildlife population. Since most accidents take place at night, recent night vision driver assistance systems are supporting the driver by automatically detecting animals on infrared camera imagery. After detecting an animal on the roadside, the orientation towards the road can give a first cue for an upcoming trajectory prediction. This paper describes an novel classification-based scheme for nighttime animal orientation estimation from single infrared images. Our system classifies already detected animals, in particular deer, as being either oriented left, right or back/front. We propose an approach based on Convolutional Neural Networks which learns multiple stages of invariant features in a supervised end-to-end fashion. Experiments show that our method outperforms baseline methods like HOG/SVM or boosted Haar-stumps on this task.
Article
The behavior of white-tailed deer (Odocoileus virginianus) and the effect of Swareflex Wildlife Highway Warning Reflectors were studied along a busy two-lane highway in Crab Orchard National Wildlife Refuge, Illinois, U.S.A. Focal-animal observational sampling was conducted through all seasons and for 24 h day−1. Deer roadside activity was primarily between 17:00 and 07:00 h and appeared unaffected by weather conditions or traffic volume. During the baseline studies of Area I (bordered on one side by a corn field and on the other by hardwood forest), 70% of the observed deer crossed or attempted to cross the highway; feeding on the grassy right-of-way was observed in 32% of the sightings, mostly after corn was depleted. In Area II (bordered on both sides by hardwood forest), 36% of the observed deer crossed or attempted to cross the highway; 68% fed on the grassy right-of-way. In the presence of traffic, deer commonly looked up from feeding and stared at the approaching vehicle. Crossing behavior was similar in both areas; deer generally walked to the highway and stopped at the edge of the pavement before moving across. Adult females usually crossed in a relaxed, cautious manner. Fawns appeared less cautious and followed adult females. Adult and yearling males (rarely observed along the roadway) crossed with more excitement than adult females and fawns. After baseline data had been obtained, Swareflex reflectors were installed for 14 months in Area I while observational sampling continued. Deer roadkills occurring between dusk and dawn were the same with reflectors deployed as during the 2 years prior to testing, even though the deer population had decreased. Behavior of deer crossing the road did not appear to be altered by the reflectors.
Article
Survival of white-tailed deer (Odocoileus virginianus) fawns has been quantified throughout much of North America. However, few studies have assessed the influence of intrinsic factors (e.g., fawn age and birth mass) and habitat on fawn survival. During 2002-2004, we captured and radiocollared 166 fawns in southern Illinois, USA, to estimate survival rates, determine causes of mortality, and identify factors influencing fawn survival. We used a known fates model in program MARK to estimate survival rates and compare explanatory models based on Akaike's Information Criterion corrected for small sample sizes (AICc). We developed 2 candidate sets of a priori models to quantify factors influencing fawn survival: model set 1 included intrinsic factors and model set 2 focused on habitat variables. We recorded 64 mortalities and the overall survival rate was 0.59 (95% CI = 0.51-0.68). Predation was the leading source of mortality (64%) and coyotes (Canis latrans) were the most prominent predator. For model set 1, model {Sage X year} had the lowest AICc value suggesting that the age at mortality varied among capture years. For model set 2, model {Slandscape+forest} had the lowest AICc value and indicated that areas inhabited by surviving fawns were characterized by a few large (i.e., > 5 ha) irregular forest patches adjacent to several small nonforest patches, and survival areas also contained more edge habitat than mortality areas. Due to the magnitude of coyote predation, survival areas could have represented landscapes where coyotes were less effective at locating and capturing fawns when compared to mortality areas. This study was the first account of macrohabitat characteristics directly influencing fawn survival. Wildlife managers can use this information to determine how habitat management activities may affect deer populations.
Conference Paper
A hierarchical framework based on Model Predictive Control (MPC) for autonomous vehicles is presented. We formulate a predictive control problem in order to best follow a given path by controlling the front steering angle while fulfilling various physical and design constraints. We start from the low-level active steering-controller presented in [3], [9] and integrate it with a high level trajectory planner. At both levels MPC design is used. At the high-level, a trajectory is computed on-line, in a receding horizon fashion, based on a simplified point-mass vehicle model. At the low- level a MPC controller computes the vehicle inputs in order to best follow the desired trajectory based on detailed nonlinear vehicle model. This article presents the approach, the method for implementing it, and successful preliminary simulative results on slippery roads at high entry speed.
Article
Motor vehicle collisions with white-tailed deer (Odocoileus virginianus) present several problems including danger to humans, vehicle damage, and deer mortality. Knowledge of factors influencing deer movements onto or across roads and highways may reduce deer/vehicle collisions on existing roads, and improve planning for future roads. We used remotely sensed data to determine characteristics associated with high accident areas. Topographic features and highway construction variables considered conducive to deer/vehicle accidents were measured around high accident road segments (≥15 accidents from 1989–1993) and randomly selected control sites. Variables were measured on aerial photographs and topographic maps within a 0.8 km radius of the road segments. Landscape composition and spatial structure were quantified with the computer program fragstats, using a statewide land cover classification derived from Landsat V TM satellite imagery. A logistic regression model composed of site variables predicted that greater distance to forest cover decreased the probability of a road segment being a high deer/vehicle accident site. The presence of adjacent gullies, riparian travel corridors crossing the road, and public recreational land within the 0.8 km radius increased this probability. A model using only landscape metrics derived from satellite imagery predicted that greater landscape diversity and shorter distances between nearby forest patches increased the probability of a road segment being a high deer/vehicle accident site. Both models discriminated between high and low deer kill sites. Therefore, proactive management of negative deer/human interactions may be accomplished through remote sensing and geographical information systems.
Article
In this paper, a model predictive control (MPC) approach for controlling an active front steering system in an autonomous vehicle is presented. At each time step, a trajectory is assumed to be known over a finite horizon, and an MPC controller computes the front steering angle in order to follow the trajectory on slippery roads at the highest possible entry speed. We present two approaches with different computational complexities. In the first approach, we formulate the MPC problem by using a nonlinear vehicle model. The second approach is based on successive online linearization of the vehicle model. Discussions on computational complexity and performance of the two schemes are presented. The effectiveness of the proposed MPC formulation is demonstrated by simulation and experimental tests up to 21 m/s on icy roads
Receding horizon control. Advanced Textbooks in Control and Signal Processing
  • W Kwon
  • S Han
Kwon, W., Han, S. (2005). Receding horizon control. Advanced Textbooks in Control and Signal Processing. Springer.
A software package for sequential quadratic programming. Forschungsbericht-Deutsche Forschungs-und Versuchsanstalt fur Luft-und Raumfahrt
  • D Kraft
Kraft, D., (1988). A software package for sequential quadratic programming. Forschungsbericht-Deutsche Forschungs-und Versuchsanstalt fur Luft-und Raumfahrt.