Guaranteeing Message Latencies on Controller Area Network CAN

1
Guaranteeing Message Latencies on Controller Area
Network (CAN)

• Garrett Chandler
• 27 March 2006
• EE 699

2
Outline

• Introduction
• System Model
• System Analysis
• SAE Benchmark Case
• Discussion and Conclusions

3
Introduction

• Common Automotive Industry View
• CAN good for most urgent data
• CAN unable to provide guarantees for less urgent
  data
• View Incorrect
• CAN scheduling virtually identical to real-time
  process scheduling
• Analysis can be applied almost without change to
  CAN latency problem

4
System Model CAN

• Typical CAN Stuff
• Assumptions
• Fixed Set of Messages
• Each w/
• Unique Priority
• Rate
• Max Size
• No RTR Messages
• Soft Real-Time as Background Messages
  (low-priority IDs)

• Hardware
• Intel 82527
• 15 slots
• 1 in only, 14 in or out

5
System Model Messaging

• Messages queued by software task
• Bounded time between event and queue entrance
• Queuing Jitter variability between subsequent
  queuings of message
• Period minimum time between same task invocations

6
System Model Queuing Jitter

• Jitter is the difference between the earliest and
  latest possible times a given message can be
  queued
• Jitter important for analysis
• Options exist to reduce queuing jitter

• Tm Period
• Jm Jitter

7
System Analysis

• Analysis will bound worst-case latency of given
  message
• Almost direct application of processor scheduling
  theory
• Assumptions
• Deadline of a message must not be more than the
  period of a message
• Controller always sends highest priority message

8
System Analysis Notation

• Deadline, Dm Maximum time between event and
  response
• Period, Tm Minimum time between events
• Worst-Case Response Time, Rm Longest time
  between message entering queue and the latest
  said message arrives at receiving nodes
• Queuing Jitter, Jm - Variability between
  subsequent queuings of message
• Queuing Delay, wm Longest time a message may
  sit in a queue before being sent
• Transmission Time, Cm Time it takes to send
  given message on the bus

9
System Analysis Response

• Rm Worst-case response time of message
• Jm Queuing Jitter
• wm Queuing Delay
• Cm Transmission Time

10
System Analysis Transmission

• Cm Transmission Time

• Accounts For
• Frame Overhead
• Data Contents
• Stuff Bits

• sm size of message in bytes
• tbit bit time of the bus (1us at 1MHz)

11
System Analysis Queuing

• wm Queuing Delay

• Set hp(m) contains all messages in system of
  higher priority than m.

• Bm is the longest time that m can be delayed by
  lower priority messages.

12
System Analysis Delay

• Bm Arbitration Delay

• Set lp(m) contains all messages in system of
  lower priority than m.

• In case of unbounded number of lower-priority
  messages of indeterminate size, Bm is 130
  bit-times.

13
System Analysis Scheduling

• Given equations assume nothing about system
  choice of message priorities
• Work on processor scheduling shows that optimal
  ordering of priorities is deadline monotonic
• Tasks with smallest difference in deadline and
  jitter should have highest priority

14
SAE Benchmark Case

• SAE Example System
• Battery, Vehicle Controller, Motor Controller,
  Instrument Panel, Driver Inputs, Brakes,
  Transmission
• 53 Messages according to Kopetz 6

15
SAE Benchmark Latencies

• Messages Ordered by D J
• 1 to 1 ID to Data Assignment

• Bold Sporadic Signal
• X failed to meet latency requirements

16
SAE Benchmark Utilization

• System can not be scheduled at 125 kbps

• Message Utilization
• Data bits / Bus bits

• Bus Utilization
• Total bits / Bus bits available

17
82527 Limitation

• Limitation
• Vehicle controller transmits more than 14 message
  IDs
• Solution
• Concatenate more data into single packet
• Results

18
Discussion and Conclusions

• Scheduling analysis developed for fixed priority
  scheduling can be adapted for use with CAN.
• Analysis applied to benchmark system.
• Paper omits discussion and analysis of
  re-transmission of messages after a detected
  error.