With the objective to improve road safety, the automotive industry is moving toward more “intelligent” vehicles. One of the major challenges is to detect dangerous situations and react accordingly in order to avoid or mitigate accidents. This requires predicting the likely evolution of the current traffic situation, and assessing how dangerous that future situation might be. This paper is a survey of existing methods for motion prediction and risk assessment for intelligent vehicles. The proposed classification is based on the semantics used to define motion and risk. We point out the tradeoff between model completeness and real-time constraints, and the fact that the choice of a risk assessment method is influenced by the selected motion model.

/pdf/a-survey-on-motion-prediction-and-risk-assessment-for-39io9df2mm.pdf

A survey on motion prediction and risk assessment for intelligent vehicles

In order to drive safely and efficiently on public roads, autonomous vehicles will have to understand the intentions of surrounding vehicles, and adapt their own behavior accordingly. If experienced human drivers are generally good at inferring other vehicles' motion up to a few seconds in the future, most current Advanced Driving Assistance Systems (ADAS) are unable to perform such medium-term forecasts, and are usually limited to high-likelihood situations such as emergency braking. In this article, we present a first step towards consistent trajectory prediction by introducing a long short-term memory (LSTM) neural network, which is capable of accurately predicting future longitudinal and lateral trajectories for vehicles on highway. Unlike previous work focusing on a low number of trajectories collected from a few drivers, our network was trained and validated on the NGSIM US-101 dataset, which contains a total of 800 hours of recorded trajectories in various traffic densities, representing more than 6000 individual drivers.

https://hal-mines-paristech.archives-ouvertes.fr/hal-01691832/document

An LSTM network for highway trajectory prediction

The safety of the planned paths of autonomous cars with respect to the movement of other traffic participants is considered. Therefore, the stochastic occupancy of the road by other vehicles is predicted. The prediction considers uncertainties originating from the measurements and the possible behaviors of other traffic participants. In addition, the interaction of traffic participants, as well as the limitation of driving maneuvers due to the road geometry, is considered. The result of the presented approach is the probability of a crash for a specific trajectory of the autonomous car. The presented approach is efficient as most of the intensive computations are performed offline, which results in a lean online algorithm for real-time application.

/pdf/model-based-probabilistic-collision-detection-in-autonomous-4g3c82t3rf.pdf

Model-Based Probabilistic Collision Detection in Autonomous Driving

This thesis is about the safety verification of dynamical systems using reachability analysis. Novel solutions have been developed for classical reachability analysis, stochastic reachability analysis, and their application to the safety assessment of autonomous cars.

Classical reachability analysis computes the set of states that can be reached by a system. If the reachable set does not intersect any set of unsafe states, the safety of the system is guaranteed. Algorithms for this problem have been developed for linear, nonlinear, and hybrid systems.

Stochastic reachability analysis measures the probability of reaching an unsafe set. One pursued approach computes over-approximative solutions for linear systems; another one generates a Markov chain which approximately computes the stochastic reachable set of arbitrary dynamics.

https://mediatum.ub.tum.de/doc/1287517/file.pdf

Reachability Analysis and its Application to the Safety Assessment of Autonomous Cars

More and more vehicles are being equipped with Automatic Emergency Braking (AEB) systems. These systems intend to help the driver avoid or mitigate accidents by automatically applying the brakes prior to an accident. Initially only rear-end collision were addressed but over time more accident types are incorporated and brakes are applied earlier and stronger, in order to increase the velocity reduction before the accident occurs. This paper describes one of the latest AEB systems called Collision Warning with Full Auto Brake and Pedestrian Detection (CWAB-PD). It helps the driver with avoiding both rear-end and pedestrian accidents by providing a warning and, if necessary, automatic braking using full braking power. A limited set of accident scenarios is selected to illustrate the theoretical and practical performance of this system. It is shown that the CWAB-PD system can avoid accidents up to 35 km/h and can mitigate accidents achieving an impact speed reduction of 35 km/h. To the best of the authors knowledge CWAB-PD is the only system on the market that automatically can avoid accidents with pedestrians.

Collision Warning with Full Auto Brake and Pedestrian Detection - a practical example of Automatic Emergency Braking

This paper deals with the problem of decision making in the context of forward collision mitigation system design. The authors present a multilevel collision mitigation (CM) approach that allows a flexible tradeoff between potential benefit and the risk associated with driver acceptability and product liability. Due to its practical relevance, algorithms that allow for an efficient incorporation of both sensor and prediction uncertainties are further outlined. The performance tradeoffs that come along with different parameterizations are investigated by means of stochastic simulations on three dangerous traffic situations, namely 1) rear-end collisions due to an unexpected braking, 2) cutting-in vehicles, and 3) crossing traffic at intersections. The results show that an overly conservative CM system sacrifices much of its potential benefit. However, it is pointed out that the vision of accident-free driving can be achieved only through cooperative driving strategies

A Multilevel Collision Mitigation Approach—Its Situation Assessment, Decision Making, and Performance Tradeoffs

Risk estimation for the current traffic situation is crucial for safe autonomous driving systems. One part of the uncertainty in risk estimation is the behavior of the surrounding traffic participants. In this paper we focus on highway scenarios, where possible behaviors consist of a change in acceleration and lane change maneuvers. We present a novel approach for the recognition of lane change intentions of traffic participants. Our novel approach is an extension of the Naive Bayesian approach and results in a generative model. It builds on the relations to the directly surrounding vehicles and to the static traffic environment. We obtain the conditional probabilities of all relevant features using Gaussian mixtures with a flexible number of components. We systematically reduce the number of features by selecting the most powerful ones. Furthermore we investigate the predictive power of each feature with respect to the time before a lane change event. In a large scale experiment on real world data with over 160.781 samples collected on a test drive of 1100km we trained and validated our intention prediction model and achieved a significant improvement in the recognition performance of lane change intentions compared to current state of the art methods.

A lane change detection approach using feature ranking with maximized predictive power

In this paper, we present the concept of a collision prevention assistant, a system that we believe can significantly contribute to road safety. We propose a new situation assessment algorithm which is tailored to the action of braking and that further accounts for the nonlinearities that arise when vehicles cut out or come to a standstill. The effect of sensor uncertainty on the performance of the proposed algorithm is modelled using a Markov chain and analyzed by means of a Monte Carlo simulation on a typical traffic situation.

Situation assessment algorithm for a collision prevention assistant

In a method and apparatus for collision avoidance or collision mitigation for a vehicle, an existing driving space between the vehicle and a potential collision object is detected. A first warning function and/or an information function is activated when a first threshold value is reached. A system intervention with autonomous partial braking combined with at least one further safety measure is activated when a second threshold value is reached.

Method and Apparatus for Avoiding or Mitigating Vehicle Collisions

This paper deals with the problem of decision making in the context of forward collision mitigation system design. We present a multi-level collision mitigation approach which allows a flexible tradeoff between potential benefit and the risk associated with driver acceptability and product liability. Due to its practical relevance, we further outline algorithms that allow for an efficient incorporation of both sensor and prediction uncertainty. The control behavior of the system is illustrated on a simulated example of a cutting-in vehicle.

J. Hillenbrand

Papers

A Multilevel Collision Mitigation Approach—Its Situation Assessment, Decision Making, and Performance Tradeoffs

A lane change detection approach using feature ranking with maximized predictive power

Situation assessment algorithm for a collision prevention assistant

Method and Apparatus for Avoiding or Mitigating Vehicle Collisions

Efficient Decision Making for a Multi-Level Collision Mitigation System