Combining behavior and situation information for reliably estimating multiple intentions

Abstract

Intersections are the most accident-prone spots in the road network. In order to assist the driver in complex urban intersection situations, an ADAS will be required not only to recognize current but also to anticipate future maneuvers of the involved road users. Current approaches for intention estimation focus mainly on discerning only two intentions based on a vehicle's behavior. We argue that for distinguishing between more than two intentions not just a vehicle's kinematic behavior but also its driving situation needs to be taken into account. In our system we estimate four different intentions by modeling and recognizing driving situations in a Bayesian Network and using the behavior as additional evidence. For the behavior based estimation we present a newly engineered feature, the Anticipated Velocity at Stop line, that turned out to be a very strong indicator for the intention. Our system is evaluated on a real-world data set comprising approaches to seven different intersections on which we can show that our approach is able to estimate a driver's intention with a high accuracy.

A survey on motion prediction and risk assessment for intelligent vehicles

“…Context and heuristics can be used to determine what maneuvers are likely to be performed in the near future in a deterministic manner [39]. For classifying maneuvers in more complex scenarios, discriminative learning algorithms are very popular, such as Multi-Layer Perceptrons (MLP) [28] Logistic regression [40], Relevance Vector Machines (RVM) [41], or Support Vector Machines (SVM) [42][43][44]. An equally popular alternative is to break down each maneuver into a chain of consecutive events and to represent this sequence of events using a Hidden Markov Model (HMM).…”

Bayesian environment representation, prediction, and criticality assessment for driver assistance systems

“…• State machines [102,185,215,239] • Fuzzy theory [102,216] • Static BNs [122,128,147,198,216,228] • DBNs [3,58,88] -HMMs [28,142,169] -Jump Markov Models [268] • Dempster-Shafer theory [185,191,250] • Special kinds of logics [97,103,111] • Various classifiers [39,95,125] In the following, representatives of the different groups are explained to make the diversity of the methods graspable. Methods trying to infer driver intentions without any physical vehicle motion indication, e.g.…”

Multi-classification of Driver Intentions in Yielding Scenarios

“…However, this serves mostly as a way to avoid transient false detections and a separate threshold variable is fit in order to perform classification, while the quality of the estimated probability is not evaluated. Bayesian network models, which give probability estimates of intentions, have been explored by [6], [7] and others. These types of probabilistic models have the potential to explicitly encode domain knowledge but often one needs to include classification type models as local distributions in the network which are similar to the algorithms evaluated in this paper.…”

“…These types of probabilistic models have the potential to explicitly encode domain knowledge but often one needs to include classification type models as local distributions in the network which are similar to the algorithms evaluated in this paper. In order to discriminate speed profiles of Stop and Go classes a likelihood function was created in [6] that compared the speeds of test examples with speeds of a data set of training trajectories while in [7] a logistic regression component fitted to a training set was used in order to estimate the probability of a subset of the possible intentions.…”