A survey on motion prediction and risk assessment for intelligent vehicles

Lefèvre, Stéphanie; Vasquez, Dizan; Laugier, Christian · 2014 · OpenAlex-citations

DOI: 10.1186/s40648-014-0001-z

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This review paper addresses the critical challenge of motion prediction and risk assessment for intelligent vehicles, specifically Advanced Driver Assistance Systems (ADAS) and autonomous cars. The authors argue that improving road safety requires mathematical models capable of predicting the evolution of traffic situations and assessing the associated danger. The paper surveys existing methods, classifying them based on the semantics used to define motion and risk, while highlighting the trade-off between model completeness and real-time computational constraints. The authors categorize motion prediction models into three levels of abstraction. First, physics-based models rely on dynamic or kinematic laws, using techniques like single trajectory simulation, Gaussian noise simulation (often via Kalman Filters), or Monte Carlo methods. These are computationally efficient but limited to short-term predictions (less than one second) as they ignore driver intent and external interactions. Second, maneuver-based models assume vehicles act independently, predicting motion by recognizing driver intent (e.g., turning, stopping) through prototype trajectories or intention estimation algorithms like Hidden Markov Models (HMMs) and Support Vector Machines (SVMs). These allow for longer-term predictions but fail to account for inter-vehicle dependencies. Third, interaction-aware models incorporate the mutual influence between vehicles, using Coupled HMMs or Dynamic Bayesian Networks, though these are rare in literature due to high computational complexity. Regarding risk assessment, the survey distinguishes between two approaches: those evaluating the risk of physical collisions and those assessing the risk of vehicles behaving differently than expected given the context, such as violating traffic rules. The authors note that the choice of risk assessment method is heavily influenced by the selected motion model. For instance, physics-based models are often paired with collision-based risk metrics, while maneuver-based models facilitate context-aware risk evaluation. The significance of this work lies in its structured classification of a fragmented field, providing a clear overview of the limitations and applicability of various prediction techniques. The authors conclude that while physics-based models are suitable for immediate reaction, maneuver-based and interaction-aware models are necessary for anticipating complex traffic scenarios. However, they emphasize that current interaction-aware models are underdeveloped and computationally expensive, identifying the integration of social interactions into efficient, real-time prediction frameworks as a key area for future research.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• anticipation
• telematics crash prediction
• situational awareness

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Theoretical Contribution: computational model