Intelligent vehicles

Jun 30, 2014 - Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard became. - 2. IMARA. Informatique,. Mathématiques,. Automatique,.
2MB taille 1 téléchargements 155 vues
Intelligent vehicles: integration and issues Z. Alsayed, G. Bresson, P. Merdrignac, P. Morignot, F. Nashashibi, E. Pollard, G. Trehard

IMARA Informatique, Mathématiques, Automatique, pour la Route Automatisée

became

RITS Robotics & Intelligent Transportation Systems

Praxitèle

Imara

Rits

1990

2002

2014

Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

-2

Robotics vs. Artificial Intelligence (AI) Robotics: science of perceiving and manipulating the physical world through computer-controlled mechanical devices

AI: any computer program which would be said "intelligent" if the same observed behavior would be so qualified when performed by a human.

Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

3

Robotics • “ Robotics is the science of perceiving and manipulating the physical world through computer-controlled mechanical devices.” [S. Thrun, 2006]

Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

4

An attempt to define an intelligent robot • Ultimate goal: ensure its survival in its environment • Ensure its energy independence • Diagnose its own state • Evaluate its perception abilities

• Achieve a mission • React properly to an unknown/abnormal situation • Learn from experience

Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

5

Work in progress with intelligent cars • Ultimate goal: ensure its survival in its environment • Ensure its energy independence • Diagnose its own state • Evaluate its perception abilities

• Achieve a mission: move safely from a point A to B • React properly to an unknown/abnormal situation • Learn from experience

Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

6

Inputs

Automation loop Process

Sensors Exteroceptive

Knowledge base

Camera LIDAR RADAR

Perception

Odometry Inertial GPS

Communications

Supervision

Proprioceptive

Planning Global Planning (Route Planning)

Local Planning (Trajectory Planning + Trajectory Coordination)

Outputs

HMI Actuators Steering/Direction

Control (Trajectory Execution)

Speed/Propulsion

How to introduce intelligence (human behavior) into the driving process? Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

7

How do we introduce intelligence into the driving process? 1. Supervision through Software Architectures: a unified framework 2. Dealing with uncertainty 3. Increase the perception using communication 4. Limit the combination explosion

Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

8

Supervision through software architectures: a unified framework

Inputs

• reaching a full autonomous driving mode in all situations impossible • self-assessment for the vehicle of its own perception abilities Process

Sensors Exteroceptive

Know. base

Camera LIDAR RADAR

Odometry Inertial GPS

Outputs

Communications

Supervision

Perception Proprioceptive

Planning Global Planning

Local Planning

HMI Actuators Steering/Direction

Control

Speed/Propulsion

E. Pollard et al., An Ontology-based Model to Determine the Automation Level of an Automated Vehicle for Co-Driving Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

9

E. Pollard, An Ontology-based Model to Determine the Automation Level of an Automated Vehicle for Co-Driving

Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

10

Longitudinal control layer Levels of automation in terms of decisions to make about…

0: fully driving

P1

Long1: Cruise

Longitudinal control

control

Long2: Dynamic Set Speed Type

P2

Long3: Autonomous CC

P3

Long4: Stop&Go

CLong: Cooperative cruise control communication

C1

Increasing needs in terms of perception and communication Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

11

Dealing with uncertainty

Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

12

Dealing with uncertainty

Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

13

Supervision: dealing with real time Supervision

Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

14

Multi-sensor fusion: • To combine properly data from multiple sensors • Deal with the problem of track spatial and temporal correlation

Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

15

Split Covariance Intersection FilterInformation Matrix Filter (SCIF-IMF)

yk0 ykn

SCIF KF SCIF

Tk0 Tkn

T2T Fusion (IMF)

Global Global Track Track

Track-to-track Fusion (SCIF-IMF)

[H. Li, et al, 2013] Track-to-Track Fusion Using Split Covariance Intersection Filter-Information Matrix Filter (SCIF-IMF) for Vehicle Surrounding Environment Perception Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

16

Perception with Transferable Belief Model

[Trehard 2014] Credibilist simultaneous localization and mapping (C-SLAM) with a lidar Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

17

Dealing with uncertainty and building generic architectures SLAM

Low level High Level

Landmark queue Drift handling

Drift estimation Absolute information (GPS, V2I…)

Loop EKF SLAM [Bresson 2013] , A General Consistent Decentralized SLAM Solution

Information from another vehicle (V2V)

Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

18

Increasing perception using communications Decorrelated maps Send

v b1 l1

Decentralized map

1

Pv Pb

vu1

Pl

b1

Receive

v

2

Pv

b1 l1

lu1 vu2

Pb

Pl

b2

Pvu1

Pb1 Pb 1

Pau1 Plu1 Pvu2 Pvu2 Pb2 Pb2

vu3

Receive

v b1

3

Pv

Pvu3 Pvu3

Pb 3

b3

Pb

Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

19

Limit the combinatorial explosion: SLAM stretching compacted grid map:

• Dedicated to “open” areas • Load in memory only the local neighborhood map slots, since the rest of the map is saved on the hard disc • a coding technique to compact and save the old or non-used far slots

Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

20

SLAM stretching compacted grid map

Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

21

Conclusion • The main issues related to autonomous vehicles can be summarized like this: • •

• •

Deal with integration problems Using redundancy and complementary information to achieve very precise state estimation Using the fact that a vehicle should know that it does not know Imitate human behavior for decision process

More experiment results on: https://team.inria.fr/rits/ Thank you for your attention! Intelligent vehicles: integration and issues – Equipe-projet RITS – E. Pollard

30/06/2014

22