Zo Computing Design

1
Zoë Computing Design
zoe

• Design Review
• December 19, 2003
• Michael Wagner (mwagner_at_cmu.edu)

2
Computing Design Drivers

• Modifiability
• Linux Environment
• Minimize Power Consumption

3
Computing Design Drivers

• Modifiability
• Year 2 field expedition is the second, not final,
  iteration of an astrobiology robot.
• Assume improvements will be made for year 3, so
  ease the effort required.
• Support technology experiments in the field.
• Linux Environment
• Minimize Power Consumption

4
Computing Design Drivers

• Modifiability
• Linux Environment
• Minimize Power Consumption

5
Computing Design Drivers

• Modifiability
• Linux Environment
• The team is experienced with Linux.
• Linux has historically had sufficient driver and
  application software support.
• Computing hardware vendors now have widespread
  Linux support.
• We can configure off-board computers identically
  to the onboard computers.
• Minimize Power Consumption

6
Computing Design Drivers

• Modifiability
• Linux Environment
• Minimize Power Consumption
• Night operations are important, so energy must be
  efficiently conserved.
• However, complex perception and planning software
  calls for high-speed computers with large
  memories.
• Rover must be able to switch components on/off.

7
Historical Lessons

• Research software can be optimized, but usually
  is not. So assume it wont be.
• The I/O buses often become overcrowded with
  sensors and devices, both expected and
  unexpected. Power switching also becomes
  impractical.
• Transfer of large data sets off the robot can be
  very time consuming.
• The console is a crutch.

8
Operating FSM
Quick Sample
Acquire Panorama
Drive to Waypoint
Hibernate
Re-plan
Plan
Plow Trench
Full Sample
9
State Description
may require significant warm-up time
10
Power Restrictions

• Need low-power (
• Autonomy CPU(s) do not face similar restrictions
• At some times, rover will collect more solar
  power than it can store in batteries
• Battery energy density high enough to support 1
  hr of night science
• Try to minimize autonomy CPU power with
  lower-clock speed modes while
• Collecting science data
• Not using compute intensive applications
  (planners, etc.)
• Possible under Linux w/ APM or APCI
• Use power relays for science instruments, buses
  and motion controllers

11
High-level Computing Design
100 Mbps Ethernet
12
High-level Computing Design
Core CPU (PMAD)
Autonomy CPU
Planning CPU
Power Relays
Battery Chargers
Power Sensors
CAN
IMU
Encoders
Digital I/O
Analog I/O
13
High-level Computing Design
Core CPU (PMAD)
Autonomy CPU
Planning CPU
LED Ctrl
Spec
Digital I/O
GPS
9-10 Firewire
Roll / Pitch
Motion Controller
RS232
14
Power Management

• Core CPU (PMAD) can switch off other components
• Autonomy CPU
• Planner CPU
• Firewire Bus 1
• Used for driving
• Nav cams, sun sensor
• Firewire Bus 2
• Used for science
• SPI cams, fluorescence camera, workspace cams

15
Core CPU (PMAD) Candidates
All core CPUs (PMADs) have PC/104 bus
interfaces, supporting expansion
16
Autonomy and Planner CPU Candidates

• Specific selection depends upon
• Planner RAM requirements
• Availability of 3U models
• Courage to use PowerPCs?

17
Next Steps

• Autonomy CPU
• Contact vendors to find power consumption numbers
• Detail RAM requirements, operating mode, CPU
  resources for
• Mission planner
• Rover executive
• Near-field stereo _at_ 4x vehicle speed
• Far-field perception / navigation system
• Create breadboard Firewire Bus
• Can we use two Firewire cards on the same bus?
• Test power switching on dual bus system

18
Next Steps

• Implement core CPU (PMAD) and start using it ASAP
  to ensure reliability
• Test CPU / Firewire hub power switching
• Implement high-rate (60 Hz) analog data
  collection for state estimation
• Implement and test data offload strategy based on
  removable hard drive
• Find an LCD flat screen / keyboard to integrate
  onto ebox
• Communications