Bayesian Perception & Decision for Intelligent Mobility

Size: px

Start display at page:

Download "Bayesian Perception & Decision for Intelligent Mobility"

Roberta Dawson
6 years ago
Views:

1 Bayesian Perception & Decision for Intelligent Mobility E-Motion & Chroma teams Inria Research Center Grenoble Rhône-Alpes Christian LAUGIER First Class Research Director at Inria San Francisco, 05/11/2015

2 Team objectives & local collaborations R&D work performed in the e-motion & Chroma Inria teams => The objective is to develop technologies for Cooperative & Human-Aware Robots in Dynamic Environments Work related to Smart Cities & Mobility Innovations : => Focusing on Perception & Decision for Intelligent Mobility in Human environments Collaborations: Toyota, Renault, Stanford & Berkeley (context Inria@SV) => Several stays (Stéphanie Lefevre, former PhD student) in Stanford & Berkeley (about 3 years) => Several Publications & Awards & Patents (Toyota, Renault, Inria-Berkeley) 05/11/2015

3 Challenge Safe & Socially Compliant Robot Navigation in Open & Dynamic Human Environments Focus on Perception & Decision under Uncertainty Mobile Robots among peoples ADAS & Autonomous Vehicles

Key Technology 1: Bayesian Perception Sensors Fusion => Mapping &

Continuous monitoring of the dynamic environment Scene

Noisy data, Incompleteness, Dynamicity, Discrete measurements +

Approach: Bayesian Perception Reasoning about Uncertainty & Time

4 Key Technology 1: Bayesian Perception Sensors Fusion => Mapping & Detection Safe navigable space (local) Embedded Perception => Continuous monitoring of the dynamic environment Scene interpretation => Using Context & Semantics Main difficulties Noisy data, Incompleteness, Dynamicity, Discrete measurements + Real time! Approach: Bayesian Perception Reasoning about Uncertainty & Time window (Past & Future events) Improving robustness using Bayesian Sensors Fusion Interpreting the dynamic scene using Semantic & Contextual information

Key Technology 2: Bayesian Decision => Understand the situation + Avoiding

Assessment Decision-making for Safe Navigation Alarm / Control Main

Real time Approach: Prediction + Risk Assessment + Bayesian Decision Reasoning

collision risk (at t+d ) Decision-making by taking into account the Predicted

5 Key Technology 2: Bayesian Decision => Understand the situation + Avoiding Pending & Future Collisions Complex dynamic situation Human Aware Situation Assessment Decision-making for Safe Navigation Alarm / Control Main difficulties Uncertainty, Partial Knowledge, World changes, Human in the loop + Real time Approach: Prediction + Risk Assessment + Bayesian Decision Reasoning about Uncertainty & Contextual Knowledge (History & Prediction) Estimating the collision risk (at t+d ) Decision-making by taking into account the Predicted behavior of the observed mobile entities (cars, cycles, pedestrians ) & the Social / Traffic rules

6 A new framework using Probabilistic Grids Patented by Inria & Probayes, Commercialized by Probayes Used by: Toyota, Denso, Probayes, IRT Nanoelec / CEA Sensing Velocity flow (particles) Bayesian Filter (each time step) 25 Hz Occupancy & Velocity Probabilities Toyota Lexus Renault Zoé Static part (Occupancy Grid) Dynamic part (Set of Particles) A Key Technology: Bayesian Occupancy Filter (BOF) Observed traffic scene Processing Dynamic Environments using DP-Grids (Occupation & Velocity Probabilities) Bayesian Inference + Probabilistic Sensor & Dynamic Models (Robust to sensing errors & occultation) Highly parallel processing (Hardware implementation : GPU, Many-core architecture, SoC)

Risky situations (Grid level, Conservative prediction) Detect

5 3 s) Risky situations are localized in Space & Time Alarm!

Urban street experiments => Almost no false alarm (car,

7 Risky situations (Grid level, Conservative prediction) Detect Risky Situations a few seconds ahead (0.5 3 s) Risky situations are localized in Space & Time Alarm! Alarm! Urban street experiments => Almost no false alarm (car, pedestrians ) Other Vehicle Mobile Dummy Ego Vehicle Crash scenario on test tracks => Almost all collisions predicted before the crash (0.5 2 s before)

8 Collision risk (Object level, Behavior-based) => Increased time horizon & complexity Trajectory prediction & Collision Risk => Patent Inria -Toyota - Probayes 2010 Intention & Expectation => Patents Inria - Renault 2012 & Inria - Berkeley 2013 Traffic Rules Risk model Intention model Expectation model

9 Main Features & Application domains Possible applications Driving Assistance & Autonomous Driving Main features Real-time & Embedded & Sensor Fusion Static & Dynamic parts characterization Mapping & Localization & Detection Probabilistic Risk for Decision-making Industrial mobile robots (AGV) Road infrastructure monitoring (Sensors & V2X) Mobility assistance (Individual or Public areas)

10 Thank you for your attention Christian.

Team-Project Geometry and Probability for motion and action

E-Motion Team-Project Geometry and Probability for motion and action INRIA Grenoble Rhône-Alpes & Laboratory of Informatics of Grenoble (LIG) Scientific leader : Christian LAUGIER (DR1 INRIA) http://emotion.inrialpes.fr