Contribution to Perception for Intelligent Vehicles - TEL - Thèses en ligne Accéder directement au contenu
Hdr Année : 2010

Contribution to Perception for Intelligent Vehicles

Olivier Aycard
  • Fonction : Auteur
  • PersonId : 835415

Résumé

Perceiving or understanding the environment surrounding of a vehicle is a very important step in building driving assistant systems or autonomous vehicles. In this thesis, we focus on using laser scanner as a main perception sensor in context of dynamic outdoor environments. To solve this problem, we have to deal with 3 main tasks: (1) identify static part and dynamic entities moving in the environment, (2) use static part of the environment to build a map of the environment and localize the vehicle inside this map: this task is know as "Simultaneous Localization And Mapping" (SLAM) and finally (3) Detect And Track Moving Objects (DATMO). Regarding SLAM, the first contribution of this research is made by a grid-based approach\footnote{An occupancy grid is a decomposition of the environment in rectangular cells where each cell contains the probability that it is occupied by an obstacle.} to solve both problems of SLAM and detection of moving objects. To correct vehicle location from odometry we introduce a new fast incremental scan matching method that works reliably in dynamic outdoor environments. After good vehicle location is estimated, the surrounding map is updated incrementally and moving objects are detected without a priori knowledge of the targets. Our second contribution is an efficient, precise and multiscale representation of 2D/3D environment. This representation is actually an extension of occupancy grid where (1) only cells corresponding to occupied part of the environment are stored and updated (2) where cells are represented by a cluster of gaussian to have a fine representation of the environment and (3) where several occupancy grids are used to store and update a multiscale representation of the environment. Regarding DATMO, we firstly present a method of simultaneous detection, classification and tracking moving objects. A model-based approach is introduced to interpret the laser measurement sequence over a sliding window of time by hypotheses of moving object trajectories. The data-driven Markov chain Monte Carlo (DDMCMC) technique is used to explore the hypothesis space and effectively find the most likely solution. An other important problem to solve regarding DATMO is the definition of an appropriate dynamic model. In practice, objects can change their dynamic behaviors over time (e.g. : stopped, moving, accelerating, etc...). To adapt to these changing behaviors, a multiple dynamic model is generally required. But, this set of dynamic models and interactions between these models are always given a priori. Our second contribution on DATMO is a method to guide in the choice of motion models and in the estimation of interactions between these motion models. The last part of this thesis reports integration of these contributions on different experimental platforms in the framework of some national and european projects. Evaluations are presented which confirm the robustness and reliability of our contributions.
Fichier principal
Vignette du fichier
hdr_Olivier_Aycard.pdf (2.51 Mo) Télécharger le fichier
Loading...

Dates et versions

tel-00545774 , version 1 (12-12-2010)

Identifiants

  • HAL Id : tel-00545774 , version 1

Citer

Olivier Aycard. Contribution to Perception for Intelligent Vehicles. Human-Computer Interaction [cs.HC]. Université de Grenoble, 2010. ⟨tel-00545774⟩
308 Consultations
280 Téléchargements

Partager

Gmail Facebook X LinkedIn More