development and debugging of fpga-based hardware-in-the...
TRANSCRIPT
© The Aerospace Corporation 2013
Does Not Contain US Export Controlled Information
Development and Debugging of FPGA-Based
Hardware-in-the-Loop Simulation Systems
Dr. Martin Panevsky Embedded System Applications Manager Zeve Akerling Advanced Degree Fellow Xingui Zhao Engineering Specialist Embedded Control Systems Department The Aerospace Corporation Flight Software Workshop
December 10-12, 2013
California Institute of Technology
2
Does Not Contain US Export Controlled Information
Outline
• Background
• FPGA System Architecture paradigm
• Modular architecture of FPGA firmware
• Implications of High-Level Synthesis Tools for VHDL and Bitfile Generation
• FPGA ASIC Emulation
• Leveraging the FPGA Technology for a Breakthrough in Avionics System
Debugging, Integration, and Insight
Modern Hardware, Enhanced Capabilities, Reduced Cost Unique Debugging and Integration Paradigm
3
Does Not Contain US Export Controlled Information
Background
• Many of the components for any Avionics HIL Simulation system
are standard and reoccurring
• The need to interface with components that are unique to a specific
avionics system leads to custom, expensive long lead hardware
• Traditionally this results in avionics HIL simulation systems that are
costly, inflexible, and time-consuming to build
• Modern COTS hardware combined with FPGA technology create a
new paradigm where most of the hardware is shifted in the COTS
direction
• The FPGA technology is the interface between COTS and custom
avionics hardware
COTS FPGA Hardware Combined With the Minimum of Custom Hardware Creates a Cost-Effective and Flexible HIL Simulation
4
Does Not Contain US Export Controlled Information
FPGA-Based HIL Evolution
COTS FPGA Hardware Combined With the Minimum of Custom Hardware Creates a Cost-Effective and Flexible HIL Simulation
• Legacy HIL has fixed custom CPU
boards and fixed custom I/O boards
• FPGA HIL has flexible custom CPU
boards, flexible COTS I/O boards,
and COTS software environment
5
Does Not Contain US Export Controlled Information
FPGA-Based HIL System Architecture
The FPGA technology is the “glue” that binds the system together
6
Does Not Contain US Export Controlled Information
FPGA Firmware Modular Design
FPGA firmware modular architecture allows for standardization and reuse of components
7
Does Not Contain US Export Controlled Information
High-Level Tools for VHDL Synthesis
• High-level COTS tools dramatically simplify the traditional challenge
of FPGA programming
• Organic modular design facilitates code reuse
High-level FPGA synthesis tools allow various subject matter experts to directly code their own designs in VHDL
8
Does Not Contain US Export Controlled Information
Applications: ASIC Emulation Using FPGA Technology
Purpose:
• Coding FPGA VHDL to ASIC specifications in order to validate and
verify functionality before or after the ASIC has been fabricated
• The VHDL proof of concept ASIC design can be directly
transitioned to an operational hardware ASIC when the VHDL code
is deployed to a radiation hardened FPGA chip
• The validated ASIC design can be used to address parts
obsolescence when implemented in a radiation hardened FPGA
• FPGA implementation of an ASIC design provides the ability to test
and stress the system by recreating almost any conceivable
hardware anomaly in the FPGA firmware
ASIC FPGA Emulation: Unique debugging capability and full visibility of the System-Under-Test
9
Does Not Contain US Export Controlled Information
FPGA-Based System Data Logger for Debugging,
Anomaly Detection and Prediction
FPGA-Based HIL enables a new paradigm of system integration and debug
10
Does Not Contain US Export Controlled Information
FPGA-Based System Data Logger for Debugging
System-level avionics tracing and debugging of entire missions
• System logic analyzer built into the FPGA logic
• Every CPU transaction, bus communication, and hardware interaction is
recorded and archived to disk in real-time for the duration of the mission
• Post-processing and analyzing the data is invaluable for system
debugging
• Ability to record and discern rare glitches that happen only once a mission
0
0.5
1
1.5
2
2.5
11586.0
808
11586.0
846
11586.0
893
11586.0
933
11586.0
969
11586.1
022
11586.1
071
11586.1
119
11586.1
158
11586.1
194
11586.1
233
11586.1
271
11586.1
319
11586.1
358
11586.1
394
11586.1
433
11586.1
471
11586.1
508
11586.1
544
11586.1
608
11586.1
646
11586.2
682
11586.2
733
11586.2
769
11586.2
808
11586.2
846
11586.2
882
11586.2
919
11586.2
958
11586.3
011
11586.3
022
11586.3
058
11586.3
093
11586.3
144
11586.3
183
11586.3
222
11586.3
271
TIME
FIFO_FULL
Glitch occurrence
extremely rare over the
duration of the mission.
11
Does Not Contain US Export Controlled Information
FPGA-Based System Data Logger for Anomaly
Detection and Prediction
FPGA-based system logic analyzer provides powerful debugging and predictive capabilities to the system developer
• Ability to capture rare events or events that happen once and
randomly in a mission – very difficult to capture with traditional tools
• Data post-processing and plotting
• Data analysis is the first step towards creating predictive models of
the system behavior
• Creating statistical regression models for anomaly and failure
prediction
• Creating statistical models for requirements violation prediction
12
Does Not Contain US Export Controlled Information
Future Work: Incorporating Elements of Artificial
Intelligence Into Avionics Fault Management
FPGA-based artificial neural networks could create a new paradigm in fault management of avionics systems
• The increasing complexity of flight software and hardware creates a
challenge for the avionics and software designers
• Trying to anticipate all the ways in which the system fails and
responds becomes increasingly difficult
• Creating deterministic software to handle all scenarios is ultimately
a dead end
• Some elements of AI incorporated into the FSW could offer a
breakthrough
• Initial simulations will focus on the use of GPUs for creating a
specialized neural network for fault management
• The ultimate goal is a FPGA-based neural network for avionics fault
management
• The FPGA implementation combines the processing parallelism of
GPUs and the relatively low power consumption of FPGAs
13
Does Not Contain US Export Controlled Information
Summary
• Shifting the balance between COTS vs. custom hardware
• Modern development tools make the FPGA technology user-friendly
• FPGA ASIC Emulation of avionics systems offers a system-level
platform for testing and debugging
• Incorporating system monitoring and diagnostics into the FPGA firmware
• Parallelism and low power consumption make the FPGA technology a
promising candidate for complex avionics systems
FPGA technology offers system-level testing and integration with COTS tools
© The Aerospace Corporation 2013
Does Not Contain US Export Controlled Information
Thank You