Tata Motors Autonomous Vehicle
Architecture
International Connected
Autonomous Vehicles
29th January 2019
Dr. Mark Tucker
Copyright, Confidential, Tata Motors Limitedc
• Introduction
• Tata/TMETC
• UK Autodrive Project and Trials• TMETC Autonomous Hexa Architecture
• Functional/Hardware/Software• Autonomous Functions
• Sensor locations
• Perceptions
• Planning - Global/Behaviour/Trajectory
• Motion Control/Drive-by-Wire Actuation • Conclusion
Overview
Copyright, Confidential, Tata Motors Limitedc
• Operations in more than 100 countries• 660,000 employees• Tata Group $100bn turnover• Tata Motors $42bn turnover
Tata Group
Tata Motors European Technical Centre
• Created 2005• Based in Coventry• Wholly-owned subsidiary of Tata Motors• Research & development principally for Tata Motors• Engineering Centre, Design Studio, Workshops• 180-strong workforce
Copyright, Confidential, Tata Motors Limitedc
• Project objectives:
• Demonstrate autonomous and connected vehicles (V2X) in real-world urban environments
• Provide insight for stakeholders including legislators, insurers and investors• Vehicles:
• Pods (RDM) in Milton Keynes
• Autonomous (TMETC and JLR)
• Connected (TMETC, JLR and Ford)• Duration
• 3 Years: November 2015 – October 2018• Funding
• Part funding from Innovate UK (around £10m of £19.2m)
UK Autodrive Project
UK Autodrive Trials - Milton Keynes October 2018
Copyright, Confidential, Tata Motors Limitedc
Sensing
Gather data from surroundings
Perception(and localisation)
Interpret sensed data
Planning
Decide what to do
Control and Actuation
Achieve Desired Trajectory
Autonomous Functional Architecture
LinuxROSC++ Python
WindowsPTP/CANMATLABSimulink
Copyright, Confidential, Tata Motors Limitedc
Autonomous Hardware Architecture
Copyright, Confidential, Tata Motors Limitedc
Open Source Robot Operating System (ROS) Bus
GPS/SLAM
Location LocationGlobal Plan
WatchdogPoint Cloud
Rviz
Trajectory Planner
Map
Locatin
1: Sensing
Gather data from surroundings
2: Perception(and localisation)
Interpret sensed data
3: Planning
Decide what to do
4: Control
Achieve Desired Trajectory
Sensors
Sen
sor
Dat
a
GPS/SLAM
Loca
tio
n
Global Planner
Glo
bal
Pat
h
Loca
tio
n
Map
Behaviour Planner
Loca
tio
n
Beh
avio
urs
Fuse
d O
bje
cts
Glo
bal
Pat
h
Trajectory Planner
Loca
tio
n
Fuse
d O
bje
cts
Beh
avio
urs
Loca
l gri
dm
ap
Traj
ecto
ry
Sensor Fusion
Fuse
d O
bje
cts
Sen
sor
Ob
ject
s
PointcloudProcessing
Loca
l gri
dm
ap
Lid
ar D
ata
Control GUI
Co
mm
and
s
All
Stat
us
System
Traj
ecto
ry
RViz
Sen
sor
Dat
a
Watchdog
All
Stat
us
StatusSpeedgoat
Motion Controller
CA
N
Drive-by-Wire
ROS Bridge
Trajectory
Status
MATLAB/Simulink
Linux/C++/Python
Traj
ecto
ry
Loca
tio
n
Supplier
Autonomous Software Architecture
Copyright, Confidential, Tata Motors Limitedc
Sensor Locations
Sensor Type
Radar
Radar
Radar
Radar
Radar
Ibeo LIDAR
Ibeo LIDAR
Ibeo LIDAR
Ibeo LIDAR
Ibeo LIDAR
Ibeo LIDAR
Velodyne LIDAR
Velodyne LIDAR
Mobileye
Camera
Camera
Camera
CameraGPS
AntennaGPS
Antenna
LIDAR RADAR Vision GPS
Copyright, Confidential, Tata Motors Limitedc
Mapping
Pointcloud processing
Mono camera object and lane detection
Sensor Fusion
ClassificationSimultaneous Localisation and Mapping (SLAM)
Localisation and Mapping (LOAM)
Perception
Copyright, Confidential, Tata Motors Limitedc
Rviz – ROS Visualisation
Copyright, Confidential, Tata Motors Limitedc
PlanningGlobal Planner• Finds the optimal route to the
destination (according to some time/cost objective)
Behaviour Planner• Static (strategic) behaviours from
the map e.g. keep in lane, stop, give way, left turn
• Dynamic (tactical) behaviours in response to the environment e.g. traffic lights and objects (and evasive trajectories to mitigate risk)
Trajectory Planner• Generate multiple obstacle free
paths• Assign speed profiles
• Select trajectory • within lane boundaries,
avoiding dynamic/stationary obstacles, comfortable (yaw rate; lateral and longitudinal acceleration and jerk; meets regulatory constraints (speed limits, stop lines, traffic lights etc)
• Select evasive trajectory
Copyright, Confidential, Tata Motors Limitedc
• Functions
• Longitudinal control – demand the throttle position and brake pressure to achieve the trajectory plan speed profile of the selected target trajectory
• Lateral control – demand steering position to follow the path of the target trajectory plan• State management –
• Return control to the driver on system errors (based on fault analysis e.g. message delays, incorrect data, drive-by-wire faults)
• Emergency or target trajectory selection• Auxiliary control – command (lights, indicators, hazards, front/rear wipers)
Motion Control/Drive-by-Wire Actuation
Motion Control
Drive-by-Wire
Copyright, Confidential, Tata Motors Limitedc
• Successful completion of trials • Pragmatic approach to Autonomous Vehicle Development
• Open source middleware – ROS
• Off-the-shelf Tools
• Linux/C++/Python/Rviz/CANalyser/MATLAB/Simulink
• Off-the shelf hardware
• Radar/lidar/GPS/IMU/cameras
• Industrial PCs/Speedgoat/Drive-by-wire mobility solution
• Bespoke third party software
• Speedgoat bridge/Sensor Fusion
• TMETC Software
• Perception/planning/motion control
Conclusion
Top Related