The BATMobile [1] is one approach to the implementation of a system capable of driving autonomously in normal highway traffic. As with human drivers, the BATMobile would be a superior driver if it were able to make predictions about the actions of other dr
Probabilistic Models of Driver Behavior Computer Science Division University of California Berkeley, CA 94720-1776, USA oza@cs.berkeley.edu
Nikunj C. Oza
The BATMobile 1] is one approach to the implementation of a system capable of driving autonomously in normal highway tra c. As with human drivers, the BATMobile would be a superior driver if it were able to make predictions about the actions of other drivers, e ectively giving it the ability to respond before such actions even occur. The ability to predict driver behavior would also be helpful in the creation of accurate models of tra c behavior used in freeway design and analysis. In this report, we discuss the creation of dynamic probabilistic networks (DPNs) that constitute models of driver behavior. Speci cally, we discuss three model structures we have created, the data used to learn the models, and how well the models predict driver behavior. We compare the three models to observe the bene ts of modeling hidden state and using deterministic variables. We also analyze the models to see what aspects of driver behavior the models have learned.
Abstract
1 Introduction The BATMobile 1] is one approach to the implementation of a system capable of driving autonomously in normal highway tra c. It uses dynamic probabilistic networks (DPNs) to model the dynamic and partiallyobservable world of driving. Such a network is to be used by the BATMobile to maintain the current belief state, where the state variables that constitute the complete state may include (for example) the locations and velocities of other cars (actual and observed values), the intentions of other drivers, and the orientation of the road. Of course, many of these variables are unobservable some or all of the time. The main aim of this research is to derive models of the factors a ecting the intentions of other drivers, such as speeding up, slowing down, and changing lanes. Creating models of driver behavior will allow an autonomous driver such as the BATMobile to anticipate potentially dangerous situations, such as another car cutting in front of it. It could also warn human drivers of dangerous situations as part of a driver assistance module. Detailed driver models are also essential in the creation of accurate models of tra c behavior used in freeway design and analysis. These also constitute models of driver preferences and behavior which can be used to assess driver competence 2]. Creating our probabilistic driver models involves identifying the variables essential to predicting human driver behavior, using a learning algorithm 3] to automatically construct a DPN model from data, and assessing how well the model predicts drivers' actions. The data comes from either a driving simulator 1] or from videotapes. From the videotapes the locations of various cars, their speeds, and other information are extracted automatically using computer vision algorithms 4]. In the Transportation community, the type of model we are construct
ing is considered to be a microscopic model because it models the behavior of individual vehicles. This is di erent from macroscopic models which describe general tra c characteristics such as density and ow 8] 16]. In this report, we rst brie y describe microscopic tra c models. In particular, we explain the motivations for developing microscopic models and survey the types of microscopic models that have been postulated and, in some cases, validated1 This previous work motivates our decision to create a DPN model. We 1
Supported by CalTrans (PATH MOU 130) and the NSF Graduate Research Training Program in Spatial Cognition. In this paper, we will limit ourselves to discussing models of highway tra c.
Probabilistic Models of ... 25页 免费 pfp5-models of governmen... 暂无评价...Organizational Behavior ... 73页 2财富值 Organizational Behavior ... 73页...
This paper develops Probabilistic Hybrid Action Models (PHAMs), a realistic causal model for predicting the behavior generated by modern concurrent percept-driv...
Probabilistic Models of ... 25页 免费 Correlation functions of... 暂无评价...Nontangential and probabilistic boundary behavior of pluriharmonic functions Let ...
暂无评价 20页 免费 Probabilistic Models of ... 25页 免费 The report of my stress ... 4页 免费 Modeling Driver Behavior... 48页 2财富值喜欢此文档...
Probabilistic Models of ... 25页 免费 SWOT of government behav... 2页 免费 Graphical Animation of B... 暂无评价 10页 免费 Understanding Behavior o.....
Learning Probabilistic Models of Link Structure Lise Getoor Computer Science Dept. University of Maryland College Park, MD 20742 GETOOR @ CS . UMD . EDU ...
Conditional random fields Probabilistic models for segmenting and labeling sequence...This behavior is demonstrated experimentally in Section 5. L? on Bottou (...
K. Improved probabilistic models for 802.11 protocol verification Abstract. The...is useful to bound the probability of such pathologically symmetric behavior....
___ Probabilistic models of the diffusion of lexical conventions in a population of agents Andrei POPESCU-BELIS ISSCO WORKING PAPER 63 July 2005 ISSCO/...
Unsupervised learning of probabilistic models for robot navigation Navigation methods for office delivery robots need to take various sources of uncertainty into...
热门文档