Embedded Bayesian Perception & Decision-making for Autonomous Mobility in Dynamic Human Environments (Invited Talk)
Résumé
This talk addresses both the socioeconomic and technical issues that are behind the development of the next generation of cars and autonomous mobile systems. After a brief overview of the main technical difficulties to be overcome, this talk will focus on the following key issue: What are the main features of an embedded perception and decisional system, specially designed to enable robots or autonomous vehicles to safely navigate in a dynamic human environment? It will first be argued that three enabling technologies are required for the development of such a system: (1) a framework for fusing multiple-sensors data in the presence of uncertainty and for interpreting the surrounding dynamic environment in real-time, (2) a method for predicting future environment changes using perception history, contextual information and some prior knowledge, and (3) a decision-making approach that has the capability to continuously evaluate the risk of future collisions and to provide maneuver recommendations for a safe navigation. It will also be shown that Bayesian approaches are mandatory for developing such technologies and for obtaining the required robustness in the presence of uncertainty and of complex dynamic situations involving human beings. This talk will be illustrated by some interesting results gathered through several R&D projects with Toyota, Renault and the French IRT (Technological Research Institute) Nanoelec.